Cloud Comp PDF

Master Thesis
Software Engineering
Thesis no: MSE-2012:82
01 2012
A Study on Cloud Computing

Security Challenges
Santosh Bulusu
Kalyan Sudia
School of Computing
Blekinge Institute of Technology
SE-371 79 Karlskrona
Sweden
This thesis is submitted to the School of Computing at Blekinge Institute of Technology
in partial fulfillment of the requirements for the degree of Master of Science in Software
Engineering. The thesis is equivalent to 20 weeks of full time studies.
Contact Information:
Author(s):
Santosh Bulusu 880216-5798
E-mail: santosh.bulusu@me.com
Kalyan Sudia 870418-0556

E-mail: kasu10@student.bth.se
University advisor(s):
Prof. Lars Lundberg
School of Computing
School of Computing
Blekinge Institute of Technology Internet : www.bth.se/com
SE-371 79 Karlskrona Phone : +46 455 38 50 00
Sweden Fax : +46 455 38 50 57
Abstract
Context: Scientific computing in the 21st century has evolved from

fixed to distributed work environment. The current trend of Cloud
Computing (CC) allows accessing business applications from anywhere
just by connecting to the Internet. Evidence shows that, switching to
CC organizations’ annual expenditure and maintenance are being re-
duced to a greater extent. However, there are several challenges that
come along with various benefits of CC. Among these include security
aspects.
Objectives: This thesis aims to identify security challenges for adapt-
ing cloud computing and their solutions from real world for the chal-
lenge that do not have any proper mitigation strategies identified
through literature review. For this the objective is to identify exist-
ing cloud computing security challenges and their solutions. Identify
the challenges that have no mitigation strategies and gather solution-
s/guidelines/practices from practitioners, for a challenge with more
references but no mitigation strategies identified (in literature).
Methods: This study presents a literature review and a snowball
sampling to identify CC security challenges and their solutions/mit-
igation strategies. The literature review is based on search in elec-
tronic databases and snowball sample is based on the primary studies
searched and selected from electronic databases. Using the challenges
and their solutions identified form literature review, challenges with
no mitigation strategies are identified. From these identified chal-
lenges with no mitigation strategies, a challenge with more references
is identified. The surveys are employed in the later stages to identify
the mitigation strategies for this challenge. Finally the results from
the survey are discussed in a narrative fashion.
Results: 43 challenges and 89 solutions are identified from litera-
ture review using snowball sampling. In addition to these mitigation
strategies few guidelines are also identified. The challenge with more
references (i.e., more articles mentioning the challenge) and no miti-
gation strategies identified is incompatibility. The responses identified
for the three insecure areas of incompatibility (i.e., interoperability,
migration and IDM integration with CC) in cloud computing security
are mostly guidelines/practices opined by experienced practitioners.
Conclusions: This study identifies cloud computing security chal-
lenges and their solutions. Where these (challenges and solutions) are
common to cloud computing applications and cannot be generalized
to either service or deployment models (viz. SaaS, PaaS, IaaS, etc.).
The study also identifies that there are methods (guidelines/practices
identified from practitioners) to provide secure interoperability, migra-
tion and integration of on-premise authentication systems with cloud
applications, but these methods are developed by individuals (practi-
tioners/organization) specific to their context. The study also identi-
fies the non-existence of global standards for any of these operations
(providing interoperability/migration/IDM integration with cloud).
This identified non-existence of global standards and guidelines
could be help academics to know the state of practice and formu-
late better methods/standards to provide secure interoperability. The
identified cloud computing security challenges (43) and solutions (89),
can be referred by practitioners to understand which areas of security
need to be concentrated while adapting/migrating to a cloud comput-
ing environment.
Keywords: Cloud computing, Security, Challenges, Solutions, In-

compatibility.
ii
Acknowledgments
At the outset we thank Dr. Lars Lundberg and Mr. Hussain Aziz for their valu-
able guidance. Mr. Hussain had given full support almost day to day interaction
for designing the thesis without which this could not be possible. We would fi-
nally want to thank friends and family for extending their support in successful
completion of thesis.
“I would first want to thank my supervisor, its a dream come true to work
with him. Every mail and every time I talked to him I learned new things. I
would thank my friends and family for their support. I thank my mother as she
gave me unstinted support all through the process. I also thank all my well wishes
who have directly and indirectly helped me to frame this work. Last but not least
I would thank the almighty for giving me this wonderful opportunity.”
-Santosh Bulusu
“I would also thank my partner, he has lots of patience and helped me a lot
when I was confused. I never hesitated him for asking help. He also put his full
effort on thesis and discussed about the problems regarding thesis, what is to be
next. Its a great pleasure working with him.”
-Kalyan Sudia
iii
Contents
Abstract i
Acknowledgments iii
List of tables vi
List of figures vii
1 Introduction 3
1.1 Aims and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.1 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Methods used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Structure of report . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Background 7
2.1 History of cloud computing . . . . . . . . . . . . . . . . . . . . . 7
2.2 Defining the term cloud computing . . . . . . . . . . . . . . . . . 7
2.3 Benefits of cloud computing . . . . . . . . . . . . . . . . . . . . . 8
2.4 Cloud Computing: Service models . . . . . . . . . . . . . . . . . . 8
2.5 Cloud computing: Deployment models . . . . . . . . . . . . . . . 10
2.6 Importance of security in cloud computing . . . . . . . . . . . . . 11
2.7 Related work in cloud computing security . . . . . . . . . . . . . 12
3 Research Design 17
3.1 Aims and objectives of study . . . . . . . . . . . . . . . . . . . . . 17
3.1.1 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Framing research questions . . . . . . . . . . . . . . . . . . . . . . 17
3.2.1 Research questions . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Research methodology . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3.1 Collecting data from literature . . . . . . . . . . . . . . . . 19
3.3.2 Collecting data from real world . . . . . . . . . . . . . . . 19
iv
3.3.3 Presenting results . . . . . . . . . . . . . . . . . . . . . . . 20
4 Literature Review 22
4.1 Process of selecting papers for study . . . . . . . . . . . . . . . . 22
4.2 Categories of Security Challenges in Cloud Computing . . . . . . 24
4.2.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2.2 Networking . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.3 Virtualization . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.4 Organization . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2.5 Scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.6 Confidentiality and Privacy . . . . . . . . . . . . . . . . . 57
4.2.7 Backup and Recovery issues . . . . . . . . . . . . . . . . . 59
4.2.8 Other concerns . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3 Identified solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4 Results of literature review . . . . . . . . . . . . . . . . . . . . . . 80
5 Surveys: Design and results 86

5.1 Survey design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.1.1 Steps for conducting an online survey . . . . . . . . . . . . 87
5.1.2 Designing the survey questionnaire . . . . . . . . . . . . . 88
5.2 Observations from survey responses . . . . . . . . . . . . . . . . . 89
5.2.1 Results: Question related to service providers, service mod-
els and cloud types used . . . . . . . . . . . . . . . . . . . 89
5.2.2 Results: Questions in relation to cloud computing incom-
patibility issue . . . . . . . . . . . . . . . . . . . . . . . . 90
5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6 Discussion 101
6.1 Validity threats . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7 Conclusions and Future Work 105

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
References 107
Appendix A Initial Set of papers 124
Appendix B Survey questionnaire 125
Appendix C Respondents 128
v
List of Tables
3.1 Research questions and Research methodologies . . . . . . . . . . 21
4.1 Business breach report blog . . . . . . . . . . . . . . . . . . . . . 31

4.2 Cloud computing challenges related to data . . . . . . . . . . . . 35
4.3 Cloud computing challenges related to network . . . . . . . . . . . 41
4.4 Virtual machine related challenges in cloud computing . . . . . . 45
4.5 Examples of cloud computing previous failure . . . . . . . . . . . 47
4.6 Organizational security concerns that challenge cloud computing . 54
4.7 Scalability related concerns which challenge cloud computing . . . 56
4.8 Challenges faced by cloud computing in relation to confidentiality
and privacy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.9 Data backup and recovery related concerns cloud computing . . . 61
4.10 Other concerns that cater security of cloud computing . . . . . . . 65
4.11 Data related challenges and solutions . . . . . . . . . . . . . . . . 81
4.12 Network related challenges and solution . . . . . . . . . . . . . . . 82
4.13 Virtulaization related challenges and solutions . . . . . . . . . . . 82
4.14 Organization related challenges and solutions . . . . . . . . . . . . 83
4.15 Scalability related challenges and solutions . . . . . . . . . . . . . 83
4.16 Confidentiality and privacy related challenges and solutions . . . . 84
4.17 Backup related challenges and solutions . . . . . . . . . . . . . . . 84
4.18 Other concerns related challenges and solutions . . . . . . . . . . 84
5.1 Compatible on-premise authentication systems used with cloud . . 99
A.1 Initial set of papers . . . . . . . . . . . . . . . . . . . . . . . . . . 124
C.1 Respondents notation . . . . . . . . . . . . . . . . . . . . . . . . . 128
vi
List of Figures
4.1 Process of selecting papers for study . . . . . . . . . . . . . . . . 23
5.1 Results of survey: Service providers used . . . . . . . . . . . . . . 89

5.2 Results of survey: Type of service used . . . . . . . . . . . . . . . 90
5.3 Results of survey: Deployment models used . . . . . . . . . . . . 90
vii
Acknowledgments 1
Acronyms used in this document
Accronym Acronym expanded

API Application Programming Interface
ACID properties Atomicity, Consistency, Isolation and Durability
AWS Amazon Web Services
Amazon EC2 Amazon Elastic Compute Cloud
Amazon S3 Amazon Simple Storage Service
ARP Address Resolution Protocol
AUP Acceptable Use Policy
AICPA American Institute of Certified Public Accountants
BC Business Continuity
CC Cloud Computing
CSP Cloud Service Provider
CSRF Cross-site request forgery
CISO Chief Information Security Officer
CICA Canadian Institute Of Chartered Accountants
CV Cumulative Voting
DR Disaster Recovery
DoS Denial of Service
DDoS Distributed Denial of Service
DNS servers Domain Name System
DHCP Dynamic Host Configuration Protocol
DBMS Database management system
ECPA Electronic Communication privacy Act
ECV Equality of Cumulative Votes
FTP File Transfer Protocol
FISMA Federal Information Security Management Act
FIPS Federal Information Processing Standard
GAPP Generally Accepted Privacy Principles
HTTP Hypertext Transfer Protocol
HMAC Hash-based Message Authentication Code
IDM Internet Download Manager
IP spoofing Internet Protocol spoofing
ISO International Organization for Standardization
LDAP Light weight Access Protocol is internets
MITM Man-In-The-Middle
MAC Media Access Control
NIST National Institute of Standards and Technology
OAUTH Open Authorization
Acknowledgments 2
Accronym Acronym expanded

PII Personal Identifiable Information
PKI Public Key Infrastructure
PCI DSS Payment Card Industry Data Security Standard
PCI QA Payment Card Industry Quality Assessors
REST Representational state transfer
SP Service Provider
SSO Single Sign On
SLA Service Level Agreement
SSL Secure Socket Layer
SSH Secure Shell
SQL Structured Query Language
SMB Small and Medium Business
SAML Security Assessment Mark-up Language
SSO Single sign-on
SPML Service Provisioning Markup Language
SOAP Simple Object Access Protocol
SAS 70 Statement on Auditing Standards
SOA Service-Oriented Architecture
TTP Trusted Third Party
TDB Technology Development Board
TCP Transmission Control Protocol
TPM Trusted Platform Module
URL Uniform Resource Locator
VMs Virtual Machines
VMM’s Virtual Machine Monitors
VPN Virtual private networks
VSG Virtual Security Gateway
VLAN’s Virtual Local Area Network
XSS Cross-site scripting
XACML Extensible Access Control Markup Language
XML Extensible Markup Language
Chapter 1
Introduction
Cloud Computing (CC) is an emerging technology that has abstruse connection

to Grid Computing (GC) paradigm and other relevant technologies such as utility
computing, distributed computing and cluster computing [55]. The aim of both
GC and CC is to achieve resource virtualization. In spite of the aim being similar,
GC and CC have significant differences. The main emphasis of GC is to achieve
maximum computing, while that of CC is to optimize the overall computing
capacity [39]. CC also provides a way to handle wide range of organizational needs
by providing dynamically scalable servers and application to work with [163].
Leading CC service providers such as Amazon, IBM, ‘Dropbox’, Apple’s ‘iCloud’,
Google’s applications, Microsoft’s ‘Azure’, etc., are able to attract normal users
through out the world. CC have introduced a new paradigm, which helps its users
to store or develop applications dynamically and access them from anywhere and
anytime just by connecting to an application using Internet [150]. Depending on
customer’s requirement CC provides easy and customizable services to access or
work with cloud applications. Based on the user requirement CC can be used to
provide platform for designing applications, infrastructure to store and work on
company’s data and also provide applications to do user’s routine tasks.
When a customer chooses to use cloud services, data stored in the local repos-
itories will be sent to a remote data center [109]. This data in remote locations
can be accessed or managed with the help of services provided by cloud service
providers. This makes clear that for a user to store or process a piece of data
in cloud, he/she needs to transmit the data to a remote server over a channel
(internet) [75]. This data processing and storage needs to be done with utmost
care to avoid data breaches.
If proper security measures are not implemented to the data transmitted and
operated on cloud, the data is at higher risk than when stored or operated in
local repositories [170]. Malicious user who wants to gain access to transmitted
data in cloud, can do that by taping into the connection between user and remote
location. He can also hack into users account and get access to sensitive infor-
mation by creating another account (by using virtualized infrastructure that CC
provides) in the same service provider with malicious intent [109]. Since cloud
computing provides different services to chose from for diversified group of users
3
Chapter 1. Introduction 4
(nave, expert, malicious etc.), possibility of having data at risk when working in
cloud computing systems is a huge.
From the above discussion, it is summarized that security challenges with
cloud adoption and cloud interoperability need to be addressed first, before imple-
menting CC in organizations [39]. A non-exhaustive search on CC challenges also
reveal that most of the organizations consider security as an important challenge
that needs to be addressed [39] [119] [90]. Even with the strongest security mea-
sures implemented there are always some other weakness found and exploited.
Henceforth, identifying security challenges, improvising and updating solutions
for handling these challenges is essential in implementing CC.
This study presents a combined effort of literature review, snowball sampling
and a survey to identify list of solutions/guidelines/practices to handle a CC se-
curity challenge that have no proper mitigation strategies defined. It will also
provide detailed challenge description and also the challenging situation faced
by practitioners (identified in literature). Models, architectures, practices and
solutions, which help to mitigate the challenges are also listed. This report has
attempted to collect, gather and present a detailed study on cloud computing
security challenges and solutions from literature. Then provide solution to the
challenge which has no identified mitigation strategies.
1.1 Aims and Objectives

1.1.1 Aim
This thesis aims to identify security challenges for adopting cloud computing and
solutions from real world for the challenge that do not have proper mitigation
strategies identified through literature review.
1.1.2 Objectives
• Identify existing cloud computing security challenges and their solutions
from literature.
• Identify the challenges that have no mitigation strategies defined.
• Collect solutions/guidelines/practices from organizations, for a challenge

with more references but no mitigation strategies proposed (identified in
literature).
• List out solutions/practices/guidelines to the cloud computing security chal-

lenge that has no mitigation strategies identified.
1.2 Research questions

1. What are the security related challenges in cloud?
2. What solutions are present to handle security related cloud challenges?
3. How does an organization handle the identified security challenge, which

has no mitigation strategies?
4. What are the solutions or guidelines to enhance security in CC where no

mitigation strategies are identified?
1.3 Methods used

Some solutions (practices, models, architectures etc.) used to handle CC security
issues are presented in academia. This paper identifies CC security issues and
solutions to handle these issues in software engineering context, with the help
of literature review (LR) and snowball sampling. A snowball sampling is done
on selected papers in the year 2009-2011 and LR is carried out between 2008-
2012. The term cloud computing was popularly being used after 2008 we chose
to perform a literature review on resources available after that year (popular CC
Service Providers (SP) such as Salesforce.com, Google Apps Engine, Rackspace
etc., came into popularity from 2008 ) [65]. Then based on results from literature,
a survey is conducted to find out solutions/practices used in organizations to the
CC security challenges with no mitigation strategies identified through literature.
The idea behind doing so is to identify all the existing solutions and best prac-
tices/solutions used in real-time to mitigate CC security challenge and bridge the
gap that stops enthusiastic users/organization from using the power of CC.
1.4 Structure of report

Chapter 2 Background, explains a brief review on terms and terminology re-
quired to understand the later sections.
Chapter 3 Research Design, explains an outline on methodologies used to infer

the expected results.
Chapter 4 Literature Review, explains detailed process followed in selecting,

sorting and describing extracted information.
Chapter 5 Survey: Design and results, explains survey design process and results
obtained from survey.
Chapter 6 Discussion, explains a discussion on findings in this report (summa-

rizes). There is also an explanation of validity threats in this section.
Chapter 7 Conclusion and Future Work, explains the results of this study and
also the future work.
Chapter 2
Background
2.1 History of cloud computing

Cloud Computing (CC) is a new term given to a technological evolution of dis-
tributed computing and grid computing. CC has been evolving over a period
of time and many companies are finding it interesting to use. Without the
development of ARPANET (Advance Research Projects Agency Network) by
J.C.R.Licklider in 1960’s and many other researchers who dreamt of improving
the interconnection of systems, CC would never have come into existence. The
advent of ARPANET, which helped to connect (for sharing, transferring, etc.)
a group of computers [86], lead to the invention of Internet (where bridging the
gap between systems became easy). This Internet helped to accelerate multifar-
ious activities such as human interaction (social media, instant messaging, etc.),
business needs of an organization (online shopping, financial services, etc.). Fur-
ther advancement in this area of Internet resulted in development of Applications
Service Provision (ASP), grid and utility computing and cloud computing [86].
CC introduced a new paradigm which changed the traditional interconnection of
systems to a pool of shared resources that can be accessed through internet.
2.2 Defining the term cloud computing

NIST (National Institute of Standards and Technology) defines cloud computing
as follows: “Cloud computing is a model for enabling ubiquitous, convenient,
on-demand network access to a shared pool of configurable computing resources
(e.g., networks, servers, storage, applications, and services) that can be rapidly
provisioned and released with minimal management effort or service provider in-
teraction” [106].
This definition clearly states that CC helps in minimizing an organization’s
expenditure towards managing resources and also reduces the burden of maintain-
ing software or hardware by its user. When burden of management, maintaining
a software/hardware is reduced, the companies expenditure and time spent to-
wards infrastructure management is reduced and time saved can be utilized in
doing some creative work. This is a huge advantage for users/organizations, which
7
Chapter 2. Background 8
not only saves time but also boosts the performance of company by saving time
spent on infrastructure.
2.3 Benefits of cloud computing

Some common benefits of CC are [119][130][3]:
• Reduced Cost: Since cloud technology is implemented incrementally (step-

by-step), it saves organizations total expenditure.
• Increased Storage: When compared to private computer systems, huge amounts

of data can be stored than usual.
• Flexibility: Compared to traditional computing methods, cloud computing

allows an entire organizational segment or portion of it to be outsourced.
• Greater mobility: Accessing information, whenever and wherever needed

unlike traditional systems (storing data in personal computers and accessing
only when near it).
• Shift of IT focus: Organizations can focus on innovation (i.e., implement-

ing new products strategies in organization) rather than worrying about
maintenance issues such as software updates or computing issues.
These benefits of cloud computing draw lot of attention from Information and
Technology Community (ITC). A survey by ITC in the year 2008, 2009 shows
that many companies and individuals are noticing that CC is proving to be helpful
when compared to traditional computing methods [123].
2.4 Cloud Computing: Service models

CC can be accessed through a set of services models. These services are designed
to exhibit certain characteristics and to satisfy the organizational requirements.
From this, a best suited service can be selected and customized for an orga-
nization’s use. Some of the common distinctions in cloud computing services
are Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS), Infrastructure-
as-a-Service (IaaS), Hardware-as-a-Service (HaaS) and Data storage-as-a-Service
(DaaS) [146][39][167][157]. Service model details are as follows:
• Software as a Service (SaaS): The service provider in this context pro-

vides capability to use one or more applications running on a cloud in-
frastructure. These applications can be accessed from various thin client
interfaces such as web browsers. A user for this service need not maintain,
manage or control the underlying cloud infrastructure (i.e. network, oper-

ating systems, storage etc.) [106]. Examples for SaaS cloud’s are Salesforce,
NetSuite [77].
• Platform as a Service (PaaS): The service provider in this context pro-

vides user resources to deploy onto cloud infrastructure, supported applica-
tions that are designed or acquired by user. A user using this service has
control over deployed applications and application hosting environment, but
has no control over infrastructure such as network, storage, servers, oper-
ating systems etc [106][130]. Examples for PaaS cloud’s are Google App
Engine, Microsoft Azure, Heroku [77].
• Infrastructure as a Service (IaaS): The consumer is provided with

power to control process, manage storage, network and other fundamen-
tal computing resources which are helpful to manage arbitrary software
and this can include operating system and applications. By using this kind
of service, user has control over operating system, storage, deployed appli-
cations and possible limited control over selected networking components
[106]. Examples for IaaS cloud’s are Eucalyptus (The Eucalyptus Open-
source Cloud-computing System), Amazon EC2, Rackspace, Nimbus [77].
• Privacy and Anonymization as a Service (PAAS): This service is

proposed as a demonstration model to provide data privacy and protection
in a particular organization. It also proposes a work-flow oriented approach
to manage data in cloud [73][157].
• Hardware as a Service (HaaS): The idea of buying a hardware or an

entire datacenter with a pay-as-you-use scheme which can scale up and
down as per user requirements can be termed as Hardware as a Service
(HaaS) [127]. Examples for HaaS cloud’s are Amazon EC2, IBM’s Blue
Cloud Project, Nimbus, Eucalyptus, Enomalism [157].
• Identity as a Service (IDaaS): This service is targeted for third party

service providers who provide Identity and access control functions (includ-
ing users life cycle and sign-on process). This can be used in combination
with various other services (software, platform or infrastructure services)
and also for public and private clouds [135].
• Data storage as a Service (DaaS): This service allows user to pay for
the amount of data storage he/she is using. With this service there is a
separate cloud formed which provides storage as a service [39]. Examples
of such kinds of users are Amazon S3, Google Bigtable, Apache Hbase, etc
[157].
• Security as a Service (SaaS): This service allows users to create their

own security policies and risk frameworks. In this kind of service cloud
users must identify, assess, measure and prioritize system risks [88].
• Anything as a Service (XaaS): This is more general form of representing
deployment of a service. These services could be of any type and ‘X’ in XaaS
can be substituted by software, hardware, infrastructure, data, business,
IT, Security, monitoring, etc. These days new service models are being
developed [127][130]. Examples are: IT as a service [96], Cloud as a Service
(CaaS) [130], Management as a Service (MaaS) [130], etc., are some other
services that are identified in literature.
2.5 Cloud computing: Deployment models

Among the service models explained above, SaaS, PaaS and IaaS are popular
among providers and users. These services can be deployed on one or more
deployment models such as, public cloud, private cloud, community cloud and
hybrid cloud to use features of cloud computing [123][9]. Each of these deployment
models are explained as follows:
• Public cloud: This type of infrastructure is made available to large in-
dustrial groups or public. These are maintained and owned by organization
selling cloud services [39].
• Private cloud: This type of cloud deployment is just kept accessible to
the organization that designs it. Private clouds can be managed by third
party or the organization itself. In this scenario, cloud servers may or may
not exist in the same place where the organization is located [39].
• Hybrid cloud: With in this deployment model there can be two or more
clouds like private, public or a community. These constituting clouds (com-
binations of clouds used, such as ‘private and public’, ‘public and commu-
nity’, etc.) remain different but yet bound together by standardized or
preparatory technology that enables application and data portability [39].
• Community cloud: This type of cloud infrastructure is shared by several
organizations and supports a specific community with shared concerns. This
can be managed by an organization or third party and can be deployed off
or in the organizational premise [39].
Usage of deployments models and services modeled provided by CC changes
how systems are connected and work is done in an organization. It adds up dy-
namically expandable nature to the applications, platforms, infrastructure or any
other resource that is ordered and used in CC [130]. This means, when an indi-
vidual/user uses more computation resources than normal, more money is spent
and when less resources are used than normal, individual/user needs to pay less
(Pay-per-use policy) [136]. This reduces organizations/individuals expenditure
towards managing resources (which includes buying computational resources, in-
stalling required software or applications to satisfy the daily computation needs
and also maintaining them) [130].
Salesforce.com (an enterprise cloud computing company) in the year 1999 was
the first to implement this idea. It provided a web based service model to sat-
isfy enterprise needs. Later in 2002, Amazon web services launched cloud-based
services, including storage, computation and even human intelligence through
Amazon Mechanical Turk [87]. Now, there are various Service Providers (SP) in
the market who offers various applications for cloud computing in various forms.
The users are provided with feature rich applications, dynamically scalable stor-
age services, application developing interfaces and many more by just signing into
a web browser/ dedicated apps. In addition to this, since CC supports remote ac-
cess feature and automatic updates (by cloud SP), any application once updated
on a site gets updated to all its users [146][174].
2.6 Importance of security in cloud computing

The power, flexibility and ease of use of CC comes with lot of security challenges.
Even though CC is a new intuitive way to access applications and make work
simple, there are a number of challenges/issues that can effect its adoption. A
non-exhaustive search in this field reveals some issues. They are: Service Level
Agreements (SLA), what to migrate, security, etc. [39]. CC has a feature of
automatic updates, which means a single change by an administrator to an appli-
cation would reflect on all its users. This advertently also leads to the conclusion
that any faults in the software are visible to a large number of users immediately,
which is a major risk for any organization with little security. It is also agreed
up on by many researchers that security is a huge concern for adoption of cloud
computing. A survey by IDC on 263 executives also shows that security is ranked
first among challenges in CC [85][7][100].
Even though a company boasts to have top class security and does not up-
date its security policies from time to time, it will be prone to security breaches
in near future. In this regard, through this detailed study, we propose to update
the readers with different distinctions (types of) in security challenges and their
solutions. We also include real-time practices to mitigate challenges, include im-
proved solutions proposed by researchers to show which areas of cloud computing
need more attention.
2.7 Related work in cloud computing security

Some methods have been proposed in literature for handling security issues in or-
ganizations implementing cloud computing. A brief discussion on these methods
is given below:
• Popovi and Hocenski, discussed security issues, requirements and challenges

that cloud service providers (CSP) need to address in cloud engineering
[119]:
1. Security issues describe the problems encountered during implementa-

tion of cloud computing (CC).
2. Security standards provide some security templates, which are manda-
tory for cloud service providers. The Open Visualization Format (OVF)
is a standard for creating new business models that help the company
to sell a product on premises, on demand, or in a hybrid deployment
model.
3. Security management models are designed based on the security stan-
dards and best practices.
• Maggi and Zanero, addressed countermeasures (anti-viruses, intrusion de-

tection systems) developed to mitigate well-known security threats. The
focus is mainly on anomaly-based approaches which are mostly suited for
modern protection tools and not for intrusion detectors. The pattern-based
changes (example: from thin client connected to the main frame or power-
ful workstations connecting to thin clients) are observed, which cause some
simultaneous changes in work environment and new problems to security of
CC [101].
• Ertaul et al., mentioned CC’s features like reduced total cost of ownership,
scalability and competitive differentiation. They claim CC also minimizes
complexity and provides faster and easier acquisition of services to cus-
tomers. Virtualization is the technique used to deal with quality of service
(QOS). Usage of CC is considered to be unsafe in an organization. For
dealing with this type of situation, they investigated a few major security
issues with CC and also existing countermeasures to those security chal-
lenges. Advantages for implementing CC from a different point of view are
also discussed. They also stated that some standards are required in CC
for security [46].
• Subashini and Kavitha, dealt with the security risks faced in the CC. They
provided empirical evidence on security risks and issues encountered dur-
ing deployment of service delivery models in an organization. The service
models are placed in cloud and the empirical validation was made in order
to justify the safety of the environment. Security was the main issue while
there were also complications with data protection and data privacy in a
continuous manner that affected the market [144].
• Md. Tanzim Khorshed et al [90] boast that cloud computing helps reduces
cost of services and improves business outcomes. But to market this and
popularize its use by IT user community, there are many security risks to
be solved. They also mentioned that the cloud services pose an attrac-
tive target to cyber attacks and criminal activities as these services have
information from many organizations and individuals stored in their repos-
itories. The author performs a survey in cloud computing to find out gaps
and security concerns and mentions 5 common types of attacks:
a. Denial of service: In this type of attack the attacker prevents the
legitimate user from accessing his resources,
b. Malicious insider attacks: This type of attack the attacker is an insider.
This person can easily gain access to sensitive user information namely:
passwords, cryptographic keys, etc.
c. Cross virtual machine side channel attacks: Is the type of attack in
which attacker resides in the same physical hardware as that of the
target virtual machine and gains access to his sensitive information ,
d. Phishing attacks: In this type of attacks the attacker sends links to
the target user through email or instant messages. These links look as
if they were sent by a trusted party but through this links the attacker
can gain access to user sensitive information,
e. Attacks targeting shared memory: The shared memory between the
user and the attacker is used to perform unwanted, unauthorized ac-
tions.
They proposed a method for automatic identification of these attacks, tested
its effectiveness by simulating attacks in a real, actual cloud setup. The
design of this model, they say, is based on machine learning models. Various
models are considered and a support vector machine (SVM) is able to detect
maximum attacks.
In the end, the authors conclude that their results are outcomes of simula-
tion and express that there can be difference in depth, volume and intensity
of attacks in simulation as compared to an actual environment. This leaves
us an impression that the method proposed still needs real world experi-
mentation and hence the attacks mentioned in this paper are still a threat
in CC environment.
• A study by Farhan Bashir Sheikh et al in [137], includes information regard-
ing vulnerable security threats from 11 articles. The authors tabulated their
findings i.e., problem discussed and technique used to solve the problem in
their paper. But in the end, they conclude expressing that cloud comput-
ing from user perspective is suffering from numerous security threats. This,
they say, is the only worth mentioning disadvantage in CC. They also list
out the following as key concerns in their point of view:
a. Users authentication: User authentication process must be improvised

to ensure that malicious users do not get access to powerful computing
systems in CC [1].
b. Leakage of data or Data loss: Data can be at risk if an unauthorized
person gains access to shared pool of resources and deletes or modifies
data. This risk can increase further if there exists no backup for that
data [1].
c. Clients trust: There must be strong authentication practices imple-
mented to ensure that the clients data is being protected from unau-
thorized access [1].
d. Malicious users handling: Malicious users can be attackers using cloud
services with a malicious intent or an insider who has gained the trust
of company but works to gain access to sensitive information stored in
cloud [1].
e. Hijacking of sessions: These kind of attacks happen when a legitimate
user is prone to phishing or insecure application interfaces that can be
exploited by attackers. Through this kind of attacks, attackers gain
user credentials and hijack legitimate users sessions [1] and
f. Wrong usage of CC and its services: Cloud computing service providers
give access to try their cloud services for a limited period of time
for free. Some users utilize this trial period to misuse the resources
obtained through CC service provider [1].
Also, in their ‘future work’ section, the authors mention that CC is not
fully mature and is a treat to both vendors and users. They also expressed
concerns of not having any strict security standards for CC. They conclude
that even though multiple tools are developed and models proposed, CC is
not secure and nothing fruitful is found.
• Iliana Iankoulova et al in [75] have performed a systematic review to identify

which security requirements need to be further researched. To find that,
the authors used an existing model with 9 sub-factors namely: access con-
trol, attack/harm detection, non-repudiation, integrity, security auditing,
physical protection, privacy and confidentiality, recovery and prosecution
to categorize their finding from 55 papers. From this review they found
that nonrepudiation, physical protection, recovery and prosecution are the
least researched in security areas. Integrity, access control and security au-
diting are the most popular areas. A surprising finding in their review is
that privacy and confidentiality had been observed only in 7% publications.
In addition to security requirements, solutions to these identified challenges
were also mentioned.
A limitation to this study is usage of an elaborated categorization scheme.
Due to this, they say, they missed including some papers (and require-
ments), which is due to selection of databases or availability. They propose
that further study in CC is required to understand why some of these re-
quirements are least researched. They also said further study should follow
another structure to describe CC security requirements (which might help
in identifying requirements missed in their study).
• Eystein Mathisen in [105] discusses some key security issues of cloud com-
puting (policy, software and hardware security) and techniques implemented
to reduce the risk. The author expresses that usage of CC will increase
in near future and more companies will share their information to cloud
servers, which could attract large groups of hackers. He also says that in
future there are possibilities for interoperability and data lock-in issues,
which can be reduced by using open standards from the time of CC adop-
tion.
The author concluded by saying that security is always addressed late while
adopting CC and also mentioned that security standards are still missing
for CC. If an organization wishes to shift to CC but is reluctant due to
lack of proper measures or standards, it can refer to Open Cloud Manifesto
which is the largest initiative surrounding open standards. These standards
are restrictive and so most companies do no wish to follow the Open Cloud
Manifesto standards.
The proposed methods address security challenges in CC and solutions to
overcome these challenges. The following points can be observed from above
related work:
1. In the study performed by Ertaul et al., he mentions that CC is considered
unsafe to be used by organizations and he also stated CC requires some
standards [9][46]. This provides a need for further research to ensure security
for all those who are using CC applications [80].
2. Eystein Mathisen concluded in their article that security is always addressed
late while adopting CC. He also say that no proper security standards for
CC exist [105].
3. Md. Tanzim Khorshed et al and Farhan Bashir Sheikh et al both advo-
cate that security challenges are still a major hindrance for adopting CC
[90][137].
4. Md. Tanzim Khorshed et al have identified some threats to CC and pro-

posed a method for automatic detection of network attacks, but it is still
not used in real world [90].
5. Iliana Iankoulova et al identified few security areas of CC to be less re-

searched and also suggest to use another way of categorization in further
studies [75].
From these studies it can be clearly understood that there are no security stan-
dards defined, even after a few researchers trying to formulate them. It can also be
understood that even though few organizations and researchers tried to formulate
strategies to handle security issues in cloud, there are still many companies that
are reluctant to join the group of CC users. Their major concern is still security
in cloud computing. This research tries to identify every possible challenge cloud
faces and their practice/solution from literature and then pick a challenge that
has no proper solutions/practices/models proposed and ask the people working
in organizations to suggest a set of guidelines/practice to mitigate/control the
challenge. This study will help both organizations and academics to identify the
extent of research. It also will help to identify a set of solutions/practices/guide-
lines to harness the power of CC securely. This study will also include benefits
of using a specific technique listed out, which can help organizations to choose a
solution that fit their requirements.
Chapter 3
Research Design
3.1 Aims and objectives of study

This thesis aims to find cloud computing security challenges and provide a set of
guidelines/solution/best practices to handle CC security challenges based on the
aims and objectives of this thesis.
3.1.1 Aim
To identify security challenges for adopting cloud computing and solutions from
real world for the challenge that do not have proper mitigation strategies identified
through literature review.
3.1.2 Objectives
• Identify existing cloud computing security challenges and their solutions
from literature.
• Identify the challenges that have no mitigation strategies defined.
• Collect solutions/guidelines/practices from organizations, for a challenge

with more references but no mitigation strategies proposed (identified in
literature).
• List out solutions/practices/guidelines to the cloud computing security chal-

lenge that has no mitigation strategies identified.
3.2 Framing research questions

From the previous section i.e., related work (discussed in section 2.7) it is not hard
to accept the fact that in the area of CC, ‘security’ is still a challenge. As long as
this new paradigm does not evolve into a more secure computing platform, which
users (organizations/individual) can trust, depend on and utilize in everyday
work, harnessing the benefit of CC is not possible. Preliminary studies show
17
Chapter 3. Research Design 18
that most of the organizations are concerned about security when considering
CC applications [9][46][51]. It is also clear from previous discussion that CC has
no specific security standards. If there are no security standards, how can cloud
be safe? Supporting this argument, another researcher mentions that CC is not
safe to be used in organizations, which means that existing solutions or practices
are not sufficient [46]. If the existing methods are not satisfactory, there could
be better ways (solutions/guidelines/practices) developed to improve security in
CC applications. Before developing better methods, existing challenges and their
solutions need to be identified (first and second research question are framed to
know challenges and mitigation strategies, specific to CC RQ 1 & RQ 2). From the
above discussion its clear that CC does not have specific security standards and
security is the major concern for adopting cloud computing [137]. To handle the
security challenges in CC, researchers have proposed, designed and experimented
different mitigation strategies (some of them discussed as related work in section
2.7). They have also framed a set of guidelines to handle security issues in a better
way. Still, there are a few security challenges unattended/insecure, which need
improved methods (observed from related work chapter 2, section 2.7). These
challenges (challenges which are still insecure/unattended and need improved
methods) can be identified from the literature review used to answer ‘RQ 1’ and
‘RQ 2’. As the next step to promote the usability and trust in security of CC,
finding out which security measures are being used by CC security professionals
in practice (for the unattended/insecure challenges) is required to be known (The
third research question is framed to investigate which security practices are used
in practice for the challenges which have no solutions RQ 3). The last section
is to list out solutions/practices/guidelines for handling CC security issues. (the
last research question helps to find out which are popular methods to enhance
the security in CC RQ 4).
3.2.1 Research questions

To find answers for issues discussed above, the following research questions are
framed:
1. What are the security related challenges in cloud?
2. What solutions are present to handle security related cloud challenges?
3. How does an organization handle the identified security challenge, which

has no mitigation strategies?
4. What are the solutions or guidelines to enhance security in CC where no

mitigation strategies are identified?
3.3 Research methodology

3.3.1 Collecting data from literature
To identify which areas of cloud computing security need more research, initially
CC challenges are found (this is done by searching the literature). Available
methods for achieving this are literature review (LR) and systematic literature
review (SLR). SLR is used to find all available data relevant to a particular
research area [30]. Since the topic of cloud computing is a novel one, literature
review and snowball sampling are employed [60].
The article written by the researchers are based on scientific papers, online
sources, journals etc. Snowball sampling revisits all these references used in
writing an article. This in turn increases the final set of papers used to understand
the problem (selected for study). In this context, since this study had threat for
finding fewer references, authors choose snowball sampling. Literature review is
also used addition to snowball sampling for collecting information to satisfy ‘RQ
1’ and ‘RQ 2’, this method helped to gather references that are missed through
snowball sampling.
Based on the information gathered from literature review, an analysis was
employed to develop general explanations. This helps researcher to identify the
key concepts, terms and also resources used by other researchers. This data is used
to develop alternative designs or find out need for further research. Literature
review using online databases involves a series of steps [52]:
1. Identifying the keywords for the topic.
2. Creating a list of possible search terms.
3. Using search engines, electronic databases to find information.
4. Modify the list of terms and repeat step 3.
Based on similarity in the findings (inferred from respective articles), identi-

fied challenges are grouped (explained in detain under data analysis section 4.2).
Literature review in this context gathers required information for answering ‘RQ
3’.
3.3.2 Collecting data from real world

Based on results of literature review and empirical study, mechanisms that are
actually used in organizations to handle the security challenges identified from
literature are needed to be found. Other methods available under empirical stud-
ies are interviews, case studies, experiment and post-mortem. The interview is
a process of gathering in-depth information to the interviewer’s question from
responder. Here the interviewer asks a question and expects an answer to that
from the responder. The interviewer can also change the questions based on re-
sponses to get required information. On the other hand surveys are referred to
research-in-the-large, their execution or measurement cannot be controlled. The
surveys generally start with a set of defined questions and target a group of peo-
ple. Case study on the other hand is an observational method where researchers
understand the on going project or activity. Experiment is a formal rigorous and
controlled investigation where normally the objective is to distinguish between
two situations for example a situation under investigation and a control situa-
tion. Post-mortem is a mix of both case study and survey. This post-mortem
contains survey features, but concentrates on a specific case [161]. The objective
here is to identify solutions/practices used in organizations. Surveys are chosen
over other methods of empirical studies, as these can be simple and also easy to
answer. The surveys are used to provide a quantitative or numeric description of
trends opinions or attitudes [36].
The surveys are used to gather the solutions/practices/guidelines to the chal-
lenges which have no mitigation strategies proposed (and have been repeatedly
cited by many papers as a challenge). The process of survey is chosen and used
to answer RQ 3, as the objective is to gather solutions/practices to a security
challenge identified through literature review. For doing this we perform a survey
on a selected group of people (who are essentially professionals working in cloud
computing and are in various designations, experience levels and companies) by
providing each of them with an online survey link which has same set of ques-
tions. Since most respondents wish to complete answering questionnaire quickly,
the process of survey will suite best in saving time and effort.
3.3.3 Presenting results

The results of literature review and surveys are going to be guidelines/practices/-
solutions. These are going to be responses from diversified group of users and with
different levels of experience. Since the observations from the literature review
are not helpful to handle the challenge identified, solutions/practices/guidelines
observed from surveys are going to be new to the reader. Hence a narrative
analysis is employed to explain the various responses observed from the surveys.
This narrative analysis will help users (individual/organizations) to consider the
possible ways to mitigate the challenge and for the researchers giving an idea as
on what are the practitioners expecting as a solution to the challenge (to answer
RQ 4).
Table 3.1: Research questions and Research methodologies
Sl no Research question Methodology em-

ployed
1 What are the security related chal- LR, Snowball sampling
lenges in cloud?
2 What solutions are present to handle Data analysis of LR and
security related cloud challenges? snowball sampling
3 How does an organization handle the Surveys
identified security challenge, which has
no mitigation strategies?
4 What are the solutions or guidelines to Narrative analysis
enhance security in CC where no miti-
gation strategies are identified?
Table 3.1 summarizes the discussion of research questions and research method-
ologies. The research questions of this study and the corresponding research
methodologies are laid out in a tabular form. As we can see, LR and Snowball
sampling are used to identify the security related challenges in cloud. Data anal-
ysis on LR and Snowball sampling helps to answer the question ‘What solutions
are present to handle security related cloud challenges?’ Surveys are used to an-
swer ‘How does an organization handle the identified security challenge, which
has no mitigation strategies?’ Finally, the survey will be analysis to find the
solutions/guidelines to enhance security in CC.
Chapter 4
Literature Review
The study starts by identifying cloud computing security challenges and their
mitigation strategies from the literature. To identify cloud computing security
challenges and their solutions, grey literature, systematic literature review, snow-
ball sampling etc., could be used, but this report uses snowball sampling and
literature review (to answer ‘RQ 1’ and ‘RQ 2’). Literature Review (LR) helps
to identify state of art in a study and snowball sampling helps to revisit into
references used in the article and find information related to the current study.
The rationale for selecting snowball sampling and literature review is as follows:
• The topic of choice (cloud computing) is new and using other techniques
might result in a few papers.
• The snowball sampling technique considers most relevant papers as initial

set and then traverses through all the references in them.
• The LR helps to identify articles that are relevant to the study but are
missed to be identified by snowball sampling.
4.1 Process of selecting papers for study

Snowball sampling starts with a set of 13 initial (will be discussed in Table A.1)
and 19 authors suggested papers (total 32 articles as initial set for snowball sam-
pling). A two level snowballing on this initial set of scientific papers (32 articles),
resulted in a total of 88 articles (32 initial + 56 articles resulting from snow-
ball sampling). The 13 initial papers were selected after an exhaustive search
on different databases (i.e., ACM, IEEE, Inspec, Springerlink, Scopus, ISI) using
a common search string (cloud computing security). Papers with more num-
ber of references, more challenges, more citation and papers published in the
period 2009-2011 were selected (as initial articles) for performing the snowball
sampling process. The articles with irrelevant abstract (not related to cloud com-
puting security), title, unavailable in full text, not related to software engineering
and published in other languages are excluded. The process is diagrammatically
22
Chapter 4. Literature Review 23
s
ecur
ity
Figure 4.1: Process of selecting papers for study
shown in figure 4.1.
In addition to the 88 papers selected through snowball sampling, 81 additional

papers were selected through a literature review (from ACM, IEEE, Scopus, Evil-
lage and ISI). The search string used was ‘(cloud AND computing) AND (security
AND challenges)’. The exclusion criteria used were same as that of the snowball
sampling (excluded papers related to electronic commerce, health care, telecom-
munication network, mobile computing, industrial etc.) and period of search was
from 2008-2012. The idea behind exclusion criteria was to focus on software
engineering.
The objective of performing literature review is to identify challenges and
solutions proposed till date (answer for ‘RQ 1’ and ‘RQ 2’), data analysis is applied
to the final set of papers obtained through literature review. Data analysis is “The
process of systematically searching and arranging the interview transcripts, field-
notes, and other materials that you accumulate to increase your understanding
of them and to enable you to present what you have discovered to others” and it
helps in [115]:
a. Reducing the data by filtering, selecting and simplifying to transform the

raw data from different sources.
b. Reorganizing data in such a way that conclusions or a logical evidence can

be built on these results.
Data analysis can be done in various methods (qualitative, quantitative, or

a combination of qualitative and quantitative). A qualitative analysis technique
‘coding’ was employed for data analysis. This coding was done based on Perspec-
tives held by subjects, because we have coded similar information, which have
shared rules and norms as well as general points of views. Other possible qualita-
tive data analysis methods are narrative summary, thematic analysis, grounded
theory, meta-ethnography, meta-study, realistic synthesis, Miles and Huberman’s
data analysis techniques, content analysis, case survey, qualitative comparative
analysis and Bayesian meta-analysis [40]. Meta-analysis is a method for sum-
marizing empirical studies that has been advocated in many areas of behavioral
research. Narrative summary is a traditional method of summarizing data from
many areas of research [37]. Grounded theory is an inductive, theory discov-
ery methodology that allows the researcher to develop a theory based on general
features of the topic [103]. Case survey bridges the gap between surveys and
case-studies to combine their respective benefits of generalizable procedures [93].
Meta-ethnography refers to the synthesis of interpretive research [114]. Qualita-
tive comparative analysis is a combination of both qualitative and quantitative
data [40]. Since the objective here is to identify CC security challenges and solu-
tions coding technique is employed. This technique refers to a collection of data
and categorizing it based on codes designed by the author. First, regularities and
patterns from articles are searched, selected and grouped in sections, this helps to
easily analyze complex data [115]. Each of these grouped challenges are explained
in section 4.2. For example, in table 4.2 that describes data related challenges
describes the challenges identified and grouped in ‘data related’ challenges. From
this table it can be understood that ‘data security’ a challenge mentioned in
data related challenges is given code ‘D1’. Similarly a challenge ‘data locality’
mentioned in ‘data related’ challenges is given code ‘D2’.
Similarly remaining tables ‘4.3’, ‘4.6’, ‘4.7’, ‘4.4’, ‘4.8’, ‘4.9’, ‘4.10’ show codes
used, description of codes and also the terms used in grouping set of codes for
‘Networking’, ‘Organizational’, ‘Scalability’, ‘Virtual machine’, ‘Confidentiality
and Privacy’, ‘Backup and Retention’ and some ‘Other concerns’ on security
challenges of cloud computing.
4.2 Categories of Security Challenges in Cloud

Computing
Cloud computing security is the major concern and has various challenges that
need attention [165][17]. From the recent surveys on IT executives and CIO’s
conducted by IDC, it was clear that security was the highly cited (74%) challenge
in the cloud computing field [3][85]. A comparison with grid computing systems
also proves that for cloud computing security the measures are simpler and less
secure [55]. Security in cloud computing is totally based on the cloud service
provider, who is responsible for storing data and providing security [63].
Based on the data analysis process explained in the previous section, terms in
literature with similar understanding (terms such as data security, data locality
etc., are categorized in data related challenges) are grouped under 8 sections and
each section is explained in detail below. These sections are grouped based on
how they are explained and based on categorizations proposed by some authors
in their discussions.
4.2.1 Data
Data security (D1)
Information from articles that discuss about data security and data protection
are considered.
Security provided by cloud SP’s might not be highly cost effective when im-
plemented in small companies. But when two or more organizations share a
common resource there is a risk of data misuse. In such situation it is required
to secure data repositories [105]. Not only the data repositories but also data
should be secured in any stage such as storage, transit or process [92]. Since
this kind of sharing resources is prevalent in the CC scenario, protection of data
is important and is the most important challenge among other CC challenges
[170][134][109]. In shared areas to keep data secure is challenging than protecting
in a personal computer [160][57]. This problem has begun due to the introduc-
tion of new paradigm CC [107]. The author of article [164] explains how data
security effects in various service models namely SaaS, PaaS and IaaS and in the
article [70] author advocates that data security is the primary challenge for cloud
acceptance and author for [107] expresses that cloud data security is an issue to
be taken care of. For enhanced security on data repositories it is important to
provide better authentication, authorization and access control for data stored
on CC in addition to on-demand computing capability [63][28][133].
Given below are three key areas in Data security that CC refers to [169]:
1. Confidentiality: When enterprise data is stored outside organizational bound-
aries it needs to be protected from vulnerabilities. To protect data from
vulnerabilities, employees must adopt security checks to ensure that their
data stays protected from malicious attacks [10][2][58]. Few test are used to
help organizations to assess and validate, to which extent data is protected
from malicious user and they are as follows [144][26][178][154]:
(a) Cross-site scripting [XSS]

(b) Access control weaknesses
(c) OS and SQL injection flaws

(d) Cross-site request forgery [CSRF]
(e) Cookie manipulation
(f) Hidden field manipulation
(g) Insecure storage
(h) Insecure configuration
Example: With the help of Payment Card Industry Data Security Stan-
dard (PCI DSS) the data is not allowed to go outside the European Union
[144][26][154]. This can also enforce encryption on certain areas of data and
by encrypting data in this way permission is given only to specific users to
access specific areas of data [119].
2. Integrity: There is no common policy that exists for data exchange. To

maintain security on client data, thin clients are used where only few re-
sources are possible. Since only few resources are given access user are not
suggested to store any personal data such as passwords. Since passwords
are not stored on desktops, passwords cannot be stolen by anyone. Integrity
of data can be further assured by [10]:
• Using some extra features which are like unpublished API’s for securing
a particular section of data.
• Using DHCP and FTP for long time has been rendered as insecure.
3. Availability: Availability is the most problematic issue, where several com-

panies face downtime (i.e., denial of service attack) as a major issue. The
availability of a service generally depends on contract signed between client
and vendor.
Some other points that need to be highlight when it comes to data security
[28]:
• Who has rights over data (i.e., does data still belong to company?)
• If there is any other company or organization being involved (i.e., is there

involvement of any third party organization) [10][78].
• Customers using CC applications need to check, if the data provided by

cloud service providers is carried out in a lawful way or not.
• If data protection fails while data is being processed, it could result in

administrative, criminal sanctions or civil type of issues (which depends on
country controlling data). These issues may occur due to multi transfers of
data log between federated cloud providers.
• Cryptographic algorithm should be maintained well and updated regularly,

failing to do so could lead in disclosing personal data [169].
• Data is not completely protected when it is encrypted and stored. When

searching for a piece of information again in CC servers care should be
taken to retrieve information in a secured process. Traditional searches can
disclose data to other companies/individuals [156]. Not only this but also
using complex ways to encrypt can also raise issues while retrieving data
from storage [126].
Data locality (D2)

Information from articles that discuss about data locality, jurisdictional issues,
risk of seizure and loss of governance are considered.
Using CC applications or storages services questions such as “does CSP al-
low to control the data location?” arise and reason for asking this question is
explained in this section [123]. We know that in CC the data can be hosted any-
where and in most cases the customer does not know the location of his data i.e.,
the data is generally distributed over number of regions [110][85][100]. It is also
known that when the geographical location of data changes the laws governing on
that data also change. This clarifies that the user’s data (information, applica-
tions, etc.,) that is stored in cloud computing (distributed over number of regions)
is affected by the compliance and data privacy laws of that country (which ever
country user’s data is located). So it is necessary that the customer should be
informed about the location his data stored in cloud [144]. SP can provide the
location of data whenever there is a change or if the SP provide a mechanism to
track the location of data it can be very helpful for customer [110]. If the customer
shows any concerns towards the location of data they should be dealt immediately
[142]. This is because if the customer is found violating laws of certain territory
his/her data can be seized by the government. Since all the information stored
in cloud computing architecture is in a same data repository, there is always a
chance for government seizing or compromising data of another company [46].
Hence, before storing data on the cloud, users must ensure providers whether
data are stored keeping jurisdiction constraints into consideration or not. They
must also verify existing contractual commitment which can shows agreement
to local privacy requirements [86][22][119][72]. For example: Some countries like
Europe, South Africa do not let their data leave country borders, as information
is potential and sensitive in nature [144] [109].
Because of all these problems some customers also concerned that their data
should stay in the same geographical locations in which they are [7]. In [55],
the author mentions that clouds will face significant challenges to handle Cloud
applications while managing the data locality of applications and their access
patterns. In addition to this, author in [105] also concerned regarding the physical
protection that changes from data center to another data center.
Data integrity (D3)

If a system maintains integrity, its assets can be only be modified by authorized
parties or in authorized ways. This modification could be on software or hardware
entities of system [179]. Data integrity in any isolated system (with a single
database) can be maintained via database constraints and transaction. But in
a distributed environment, where databases are spread out in multiple locations
data integrity must be maintained correctly, to avoid loss of data [99][100]. For
example when the premises application is trying to access or change data on a
cloud the transaction should be complete and data integrity should be maintained
and failing to do so can cause data loss [75]. In general every transaction has
to follow ACID properties (Atomicity, Consistency, Isolation and Durability) to
preserve data integrity [179][46][130][19]. This data integrity verification is one
of the key issues in cloud data storage especially in case of an untrusted server
[120][81].
Web services face problems with transaction management frequently as it
still uses HTTP services. This HTTP service, does not support transaction or
guarantee delivery. The only way to handle this issue is by implementing trans-
action management at Application Programing Interface (API) level. There are
some standards (such as WS-Transaction and WS-Reliability) to manage data
integrity with web services exists. But since these standards are not matured
they are not implemented. The majority of vendors who deal with SaaS, expose
their web service API’s without any support for transactions. Additionally, each
SaaS application may have multiple levels of availability and SLA’s (Service Level
Agreement), complicating it further with management of transactions and data
integrity across multiple SaaS applications.
Lack of data level integrity controls could result in some profound prob-
lems. Architects and developers need to handle this carefully, to make sure that
database’s integrity is not compromised when shifting to cloud computing appli-
cations [144]. Failing to check data integrity may lead to data fabrication or in
some cases even if data is removed by CSP as its rarely accessed, the user won’t
be known until he attempts to access [153]. In article [98] the author compares
protocols that are used for remote data integrity and expresses that the proto-
cols in comparison were either focused on single server scenario or multi-server
but not in dynamic situation such as CC. In the article [126], author expresses
concern regarding maintaining a local copy to check the data integrity by each
user. CSP’s are supposed to have some user’s metadata to grant access or iden-
tify a user. To manage the integrity of data in cloud storage metadata should be
managed correctly [31].
Author in [120] has performed a comparison of different protocols (entire
data dependent tag, data block dependent tag, data independent tag based, data
replication-based protocol) to check data integrity that are discussed and com-
pared. The comparison shows most proposed methods are having data integrity
as a primary objective and also support dynamic operations in cloud storage. But
there is room for improving these methods.
Data segregation (D4)

Another issue in cloud computing is multi-tenancy. Since multi-tenancy allows
multiple users to store data on cloud servers using different built-in applications
at a time, various user’s data resides in a common place. This kind of storage
shows a possibility for data intrusion. Data can be intruded (malicious user
retrieving or hacking into others data) by using some application or injecting a
client code [144][72]. The user should ensure that data stored in the cloud should
be separated from other customer’s data [55][85][123]. Article [142] suggests that
an encryption scheme used should be assessed and certified that they are safe and
cloud provider should use only standardized encryption algorithms and protocols.
Vulnerabilities with data segregation can be detected or found out using the
following test [144]:
1. SQL injection flaws
2. Data validation
3. Insecure storage
Data access (D5)

Information from articles that discuss about data access, access rights, privileged
user access, access control, administrative access are considered.
This issue mainly relates to security policies. Policies are described as “Conditions
necessary to obtain trust, and can also prescribe actions and outcomes if certain
conditions are met” [121]. Every organization has their own security policies.
Based on these policies employee will be given access to a section of data and
in some cases employees might not be given a complete access. While giving
access it is necessary to know which piece of data is accessed by which user
[11][109]. And for this various interfaces or encryption techniques are used and
keys are shared with only authorized parties. Wrong management of keys can
also cause difficulty in providing security. To prevent wrong management of keys
access control list might be used, but with increase in the number of keys, the
complexity of managing keys also increases [95]. Even in the case of interfaces
used to manage security, if the number of interfaces increase management of access
can also become complicated [166].
In any case, to ensure that data stays away from unauthorized users the
security policies must be strictly followed [144][72]. The unauthorized access
could be from an insider or by any user trying to access CC [136]. The data
access control is a specific issue and various standalone approaches used in access
control of data in CC are mentioned in article [45].
Privileged user access: Since access is given through the Internet, giving ac-
cess to privileged users is an increasing security risk in cloud computing. When
sensitive data is transferred through the Internet there is a possibility for an
unauthorized user to gain access and control data. To avoid this, the user must
use data encryption and additional protection mechanisms like one time password
protection or multi-factor authentication, that can be used to provide strong au-
thentication and encrypted protection for all administrative traffic [3][86][22].
Data confidentiality issue (D6)

Cloud computing allows users to store their own information on remote servers,
which means content such as user data, videos, tax preparation chart etc., can be
stored with a single cloud provider or multiple cloud providers. When users store
their data in such servers, data confidentiality is a requirement [170]. Storing of
data in remote servers also arises some privacy and confidentiality issues among
individual, business, government agency, etc., some of these issues are mentioned
below [144][76]:
1. Privacy of personal information and confidentiality of business and govern-
ment information implications is significant in CC.
2. The terms of service and privacy policy established by cloud provider are the
key factors that vary user’s privacy and confidentiality risks significantly.
3. Privacy and confidentiality rights, obligations and status may change when
a user discloses information to cloud provider based on information type
and category of CC user.
4. The legal status of protections for personal or business information may be
greatly affected by disclosure and remote storage.
5. Location of information may have considerable effects on the privacy and
confidentiality information protection and also on privacy obligations for
those who process or store information.
6. Information in cloud may have multiple legal locations at the same time
but with differing legal consequences.
7. A cloud provider can examine user records for criminal activity and other
matters according to law.
8. Access to the status of information on cloud along with privacy and con-
fidentiality protections available to users would be difficult due to legal
uncertainties.
In addition to these to maintain confidentiality understanding data and its clas-

sification, users being aware of which data is stored in the cloud and what levels
of accessibility govern that piece of data should also be known [160].
Data Breaches (D7)

Since data from various users and organizations is stored in a cloud environment,
if user with malicious intent enter the cloud environment, the entire cloud en-
vironment is prone to a high value target [144][166]. A breach can occur due
to accidental transmission issues (such breaches did happen in Amazon, Google
CC’s) or due to an insider attack [100][143]. In any case of breach data is com-
promised and is always a security risk which is also a top threat mentioned by
CSA [143]. There is a high requirement for breach notification process available
in the cloud. It is because if breaches are not notified the cloud might not be able
to notify serious attacks [90].
Table 4.1: Business breach report blog
Threat Impact Resulting in

Pseudo Risk
External Crimi- Greatest (73%) Least (30,000 67,500
nals Pose compromised
records)
Insiders Pose Least (18%) Greatest 67,500
(375,000
compromised
records)
Partners are 73.39% 73.39% 73,125
middle
The business breach report blog gives information on the impact of breaches
[144], as shown in table 4.1. We can see that the threat of external criminals is
greatest by 73% but with least compromised records. On the other hand threat
of insiders is least with 18% but the impact they cause is greatest.
Reliability of data storage (D8)

As long as there exists no problems with virtualization manager, developer will
have better control over security. The virtual machines have many issues within
it, but it still a good solution for providing secure operation in CC context. With
growing virtualization in every aspect of cloud computing, there is an issue with
reliability of data storage and owner retaining control over data, regardless of its
physical location [144]. The users also think that storage mechanisms are not
reliable in CC [33]. In CC, reliability of data storage is a general issue [99]. This
issue constitutes to every data entity stored in the cloud and when Infrastructure
as a service (IaaS) is considered the service provider is expected to assure that
an organization’s data is kept secured along its life cycle (even after the user has
removed his account) [38]. Another issue to be considered here is that even a
virtual machine needs to be stored in a physical infrastructure. This can also
cause security risks, which needs to be protected [125][164]. In addition to these
problems article [126] explains various storage concerns and show to what extent
cloud can be dependable with respect to reliability of data storage.
Data center operations (D9)

Information from articles that discuss about losing control over data, incident
response, data center operational management, data center operations, data man-
agement, disaster recovery, data transfer bottlenecks, incident response are grouped
her.
Organizations using cloud computing applications are concerned about pro-
tecting data while it’s being transferred between the cloud and the business. The
concern is about what will happen to user’s data if something happens to cloud
storage? [162]. If data is not managed properly, data storage and data access
can become an issue [15]. In article [54] author expresses his concern that though
there is growing interest in cloud governance issues such as data integration, data
consistency, policy management etc., are not given required attention. Adding to
this [178] mention that cloud is not secure unless mechanisms to debug, diagnose
distributed queries and analyze exists for the cloud providers.
Explanation on issues related to data center operation are discussed below:
Recovery: In case of disaster, the cloud providers should be answerable to
users questions, such as what happened to data stored in the cloud? If the cloud
service provider does not replicate data across multiple sites, such system could
result in failure under certain circumstances (if disaster’s occur)[119][10][162].
Even though there is a disaster in cloud it should be able to recover or provide
some other means to avoid halting in the business needs of user [65]. Therefore,
service providers can be questioned “is there any option for complete restoration
if exists some way and how long does it take?” [119]. In case of PaaS, disaster
and recovery are issues that need to be paid more attention and in IaaS data
center construction is a key point [164].
Losing control over data: Data is outsourced, control of data will be lost
gradually. To avoid major loss there should be transparency in how data is
managed and how data is accessed [142]. As a solution to this, Amazon Simple
Storage Service S3 API’s, provide both object level access control and bucket
level control. Each authenticated user under this kind of security in a system is
authorized to perform certain actions, specific to each object which he/she needs
access to perform his/her task. In case of bucket level control the authorization
is granted at the level of bucket (is a container for objects stored in Amazon S3).
This security can be accessed at both object and bucket level [46].
Data transfer bottleneck: If potential consumer transfers data across the cloud
boundaries, it might lead to data transfer cost. To reduce cost, when CC appli-
cations are used cloud user and cloud providers need to focus on the implication
of data placement and data transfer at every level. During amazon development,
ship disk was used to overcome this issue with Data transfer bottleneck [13][12].
Data sanitization (D10)

Information from articles that deal about data sanitization and insecure destruc-
tion or incomplete data deletion are considered.
It is the process of removing unwanted or outdated sensitive data from the
storage device. When a user updates data in the cloud, he/she can secure the data
by encrypting while storing data on the cloud. Users are very much concerned
about what will happen to the data after it passes its user’s “use by date” date,
will it be deleted after the contract is completed? [69][46][78][7]. Even if the
data has to be deleted or no longer needed should be deleted in a secure way
such a way that unauthorized access is not possible [126][100][142]. It is also a
benefit if the user is kept informed how his data is deleted (if asked for deletion)
and also helps user keep informed if the service provider is keeping the data even
if asked for deletion [119]. Amazon Web Services (AWS) procedure includes a
decommissioning process when the storage device reached to the end of useful
life. It also means that the user data is not exposed to unauthorized users.
Sanitization is also applied to backup the data for recovery and restoration of
service [78]. While performing an anti-malware scan identifying useful piece of
data and deleting unwanted information can become complicated [138].
Example: Researchers are obtaining data from online auction and other sources.
They also extract data by retrieving large amounts of data from them (with proper
equipment, we can recover the data from the failed drives which are not disposed
by cloud providers) [78].
Data storage and lock-in (D11)

Information from articles that deal about data storage, where data is stored?,
problems with DBMS and data lock-in are considered.
Data storage is a concern in CC [54]. There are a number of data storage
concerns expressed by different authors and as follows:
• Is the data secure? Will data be available when requested? [136][92].
• Authors of [90][5][142][162][17][110] have mentioned data loss or leakage as
a challenge, concern or issue in their articles.
• Many doubts such as how data is stored, where is it stored (is it dis-
tributed in various places?), what will happen if the cloud provider is
taken over, what security measures are taken to protect the user’s data
etc., [162][119][109][45].
• Since a large amount of data is stored in cloud it can attract attention from
malicious users [166]. Due to this user might not be interested to store
mission critical data for processing or storing into CC [55]. Some others
feel that data stored in CC as insecure and not reliable [33][108].
• Customer lock-in might be attractive for service providers, but it is an

issue (viz., vulnerable to increase in price, reliability problem, etc.,) for
customers [13][12][92][110]. As an example customers of ‘Coghead’ were
forced to re-write their applications on another platform when its cloud
computing platform got shut down [35].
• It’s a known fact that different organization’s data is stored in a place and
this increases the possibility that an organization data is sold to another
organization for money. The author wants to bring it to notice that since
all the data is stored in the same place there are increased chances of data
misuse, data can be intentionally leaked and if this happens the customer
is at loss [100][133].
• The author of [142] mentions that data loss/leakage issue affects only public
cloud.
The issues with data storage in CC can arise when proper sanitization or
segregation of data is not implemented, which could result in users not able to
extract their data from repositories when necessary or when company willing to
shift data to another location.
Data storage: With cloud computing users can utilize a wide variety of flexible
online storage mechanisms to store their information, which have been known as
computing and storage “clouds”. Examples are Amazon S3, Nirvanix CloudNAS
and Microsoft SkyDrive [26]. The architecture of storage mechanism also depends
on cloud type e.g., internal or external cloud computing and types of services i.e.,
SaaS, PaaS or IaaS. This variation exists because in internal cloud computing
organization keeps all data within its own data center but in case of external
cloud computing data is outsourced to CSP [132]. In any of these cases the
data is not under the control (physical or logical state) of user and traditional
cryptography can not be used by user [68][95][164]. Since in CC data is stored
in a remote location and even traditional cryptographic algorithms cannot be
applied the security of data stored in remote locations is a huge concern in cloud
computing [153].
Users while accessing flexible storage mechanisms can maintain a local trusted
memory, use a trusted cryptographic mechanism and upload data into the cloud
(by doing this the user need not have to trust the cloud storage provider); to
verify the integrity of data user can have a short hash of in local memory and
authenticate server responses by re-calculating the hash of received data [26].
To create trust in cloud storage, data storage systems need to fulfill different
requirements such as maintaining user’s data, high availability, reliability, per-
formance, replication and data consistency; but since these requirements are all
interrelated and conflicting no system has implemented all of these at once. DaaS
providers facilitate storage as a service by implementing one feature over other,
which is mentioned by its customers in Service Level Agreements (SLA) [168].
To assure security of information, CSPs should take care to protect the data not
only when it’s stored but when its under transmission [31][123]. In article [11],
author mentions storage devices should support different storage patterns. In
addition to this various other concerns in relation to data storage specific to IaaS
are explained in [38]. To have a better security in cloud storage long term storage
correctness and remote detection of hardware failure is suggested [126].
Table 4.2: Cloud computing challenges related to data
Relation Description Code

Data Security D1
data locality D2
Data integrity D3
Data segregation D4
Data access D5
Data related Data confidentiality issue D6
Data breaches D7
Reliability of data storage D8
Data center operations D9
Data sanitization D10
Data storage D11
4.2.2 Networking
Network security (N1)
Information from articles that discuss about network security and VPN network
are considered.
Data should not be leaked while transmission and it is one of the requirements
in information security [136]. To prevent leakage of sensitive information while
transferring, a strong traffic encryption technique such as Secure Socket Layer
(SSL) and the Transport Layer Security (TLS) are required. Sensitive data are
obtained from enterprises, processed by any service application and stored at the
service vendor end. Amazon Web Services (AWS), provide more protection to its
users from traditional network based attacks like MITM (Man-In-The-Middle)
attacks, IP spoofing, port scanning, packet sniffing, etc. The assessment tests to
find vulnerabilities in security are based on the following [144]:
1. Network penetration and packet analysis
2. Session management weaknesses
3. Insecure SSL trust configuration
VPN network: If an organization is distributed globally and employees a

single vendor, then such organization might experience lower transfer rates when
sending a file from one side to another side. A solution to this is usage of Virtual
Security Gateway and maintaining multiple vendors, for implementing this usage
of some commercial solutions that give customer-controlled security in a cloud
is necessary. This helps to establish a bridge over private infrastructure, where
control over cloud lies within the organization. It enables confidentially leverage
over the cloud for redundancy, scalability and failover during critical transitions,
which may lead to scale up grow or scale down to the organization or business
[3].
Given below are some of the network attack types, concerns and suggestions
that various authors mention that can be aggregated under network security:
• The author in [14] mentions how challenging it is to identify the intrusion

severity in CC.
• Author of [90] mentions attack methods such as phishing, fraud, Denial of

Services (DoS) and account hijacking that are used to steal user credentials.
• The authors of [51] mentions that using traditional network based meth-
ods such as IPSec proxies encryption and digital signature, key exchange
through SSL proxy which are still being used in the cloud are insecure.
The authors mention vulnerabilities of using such methods in the article by
considering Amazon Web Services (AWS), Azure and Google App Engine
(GAE) as examples.
• Attack types mentioned in [169] launching dynamic attack points, key and
password cracking, hosting malicious data, botnet command and control,
DDOS, building rainbow tables and also CAPTCHA solving farms.
• The author in [134] mentions, backdoors, TCP hijacking, replay, social en-
gineering (where the attacker tries to gain private information from user’s
social behavior), dumpster diving (information from trash can be helpful),
password guessing, trojan horses and malware are some of the network at-
tacks mentioned.
• Metadata spoofing is another kind of attack where a new system similar to

cloud system can be built by analyzing and re engineering from metadata
[85].
• Account or service hijacking [17][5][110]. In addition to these Dos, IP ad-

dress modifying helps malicious users to hack into accounts [15].
• Some attack types that are specific to IaaS are DDOS, MITM, port scanning
and IP spoofing [38]. The author of [164] points out network security,
transport security as key security issues for IaaS in CC.
• Side channel attacks and incident handling [110]. SQL injection and phish-
ing by service provider [7].
• The problems account or service hijacking mentioned above affects public

and not private cloud [142].
In addition to all these kinds of attacks, the author in [75] mentions that even
the cloud firewalls are still under chaotic state. From all the attack types and
point put before us there is a strong requirement for security measures, anyone’s
credentials can be stolen easily with all these different types of security risks
pointed out. If an attacker can gain access to someone’s credentials, he or she can
eavesdrop on a customer’s activity, transactions and also become a huge threat
to user’s data [51]. A server side protection which includes application security
and virtual server should also be provided to strengthen network security [92].
Application security (N2)

Information from articles that discuss about web application security, API secu-
rity, application vulnerabilities and application security are considered.
With new advantages cloud also bring to the developers novel vulnerabilities and
threats related to APIs [5]. It is known that in the cloud, any application or
software that is used lies in cloud but not with the actual user and if this soft-
ware/application has vulnerabilities then it can have a decremental impact on all
the customers using the cloud [142][17][164]. If the insecure APIs are not secured
the vulnerabilities in that can lead to compromising security, can also lead to
men in the middle of attacks and affect the availability of CC [166][169][139]. Se-
curity surveys bring to notice that application level attacks are more bandwidth
efficient compared to network level attacks [65]. In [110] the author mentions
insecure APIs as one of the top threats to CC. Traditional security mechanisms
such as network firewalls, network intrusion detection and prevention mechanisms
do not adequately satisfy being used as a solution [70]. The typical issues arising
with applications technology are: Session riding, hijacking and injecting vulner-
abilities. Other web-application-specific vulnerabilities are browser’s front-end
components in which, data sent from the user component to server component is
manipulated [144]. The author of [31] also mentions that modern browsers com-
patible with AJAX is not secure though it’s good for I/O and data encrypted by
XML needs to be also encrypted by using transport layer security. XML signature
attacks, browser based attacks for cloud authentication are few other insecurities
in application security that can be used for cloud authentication [80]. One easy
way of access cloud for an intruder is to search for saved passwords in history
which can be a huge risk and using insecure browsers [123]. Hence securing the
browser is necessary to secure transaction between users and his application that
are stored in the cloud [92]. The application security is the main threat to SaaS
platform [164]. In article [142] author mentions that both public and private
clouds are affected by insecure applications.
Host and Network intrusion (N3)

This problem arises in PaaS, where control might be given to user by SP to some
extent. Service providers should keep in mind that control below the application
level, such as host and network intrusion shall not be given to the user and the
provider should maintain inaccessibility between applications [144]. By using
programs such as trojan horses and malware which leak sensitive data can help
intruders to gain access to sensitive information [134][80][67].
Denial of service (N4)

Denial of service attacks are possible in cloud which can be a threat to data
under transmission [32][136][169] [105]. Unlike bypassing preventive and security
measures the attacker uses methods such as packet splitting, payload mutation,
shell-code mutation and duplicate insertion [65]. The three proposed distinctions
[159][24][46] are:
• Direct denial of service: If the workload becomes high in cloud com-

puting operating system, the environment will automatically allocate more
computational power (more service instance, more virtual machines) to
overcome this additional workload. But since resources are limited with ser-
vice provider even providing additional resources for an extended amount
of time is not quite possible. In such case (flooding scenario) the cloud
system is not working against the attackers (providing more computational
power) but to some extent system is giving attackers rights to do damage on
service’s availability, starting from a single flooding attack. The attacker to
cause disturbance does not need to attack ‘n’ servers, an attacker can target
one of the cloud based address in order to attack ‘n’ servers and perform
full loss of availability [80][67].
• Indirect denial of service: In-direct denial of service, attackers’ damage

availability of a service. If an instance in cloud service is flooded with too
much work load, other services running on the same cloud server might
face issues with availability. If resources are completely utilized by a single
instance, other instance has to face consequence and wait for resources to
be allocated.
Depending on level of sophistication, if cloud service provider notices lack
of availability and tries to “evacuate”, affected instances on to other servers
the complexity rises even further. The reason for over complication is,
flooding attack which existed in previous server will now get extended to
the server where it is transferred and this situation is called “jump over”
[79][80]. In the worst case, to manage flooding attacks utilization of another
cloud computing system is suggested. In that case, cloud should provide
more computational resources, respectively fending, until it reaches the full
loss of availability [80][67].
• Distributed Denial of Service (DDoS): This is even more dangerous

and difficult to identify or control. In this type of attack the attacker floods
the target servers by controlling a large group of geographically distributed
hosts. Since Cloud is shared by a large group of users, DDoS could have
greater impact on customers [100][83].
Men in middle of attack (N5)

Attackers create an independent connection which relay on the messages between
user and provider. The attacker makes the user and provider believe that this
connection is secured and makes them talk directly, but behind the scene attacker
controls the whole connection and receives every message which is sent between
them (user and provider) [159][46][100]. The attacker can also possibly modify
the message before sending to the respondent [134]. XML signature wrapping is
one such attack where the attacker without changing the header modify the data
in it and sends to the receiver [31]. This makes clear that in cloud computing,
while data is in transfer state denial of service is a threat [32].
IP Spoofing(N6)
Using somebody’s IP address and creating TCP/IP packets is called IP spoofing.
In this scenario intruder gains access to trusted system and sends messages as
if trusted host is sending these messages. Here the malicious user impersonates
the actual user with the help of IP address [169][134]. Amazon EC2 cannot
send instance spoofed network traffic. Firewall designed in amazon web server
prohibits an instance sending messages with any other IP address or MAC address
other than its own [46].
Port scanning (N7)

A port is a place from which data goes in or out of a system. When a system’s
security fields are configured to send or receive data through a port. Then that
specific port is vulnerable to port scan. When a network is scanned for vulnerabil-
ities port scanning shows these vulnerable ports as open doors. When a computer
tries to access internet, a port is opened by default and it is not possible to stop
port scanning [130][46]. This can cause security breaches in cloud computing
platforms.
Packet Sniffing (N8)

Packet sniffing is listen to network devices and capture raw packets [169]. If a
software finds a packet suiting a particular criteria, it logs into a file (login and
password may be mandatory and significant ones). A virtual instance running in
promiscuous mode can’t receive or “sniff” traffic which is intended for a different
virtual instance. The hypervisor will not deliver any traffic when customers place
their interfaces into promiscuous mode. If two virtual instances are determined
on the same physical host and are owned by same customers then they cannot
listen to each other’s traffic. For Amazon EC2 attacks such as Address Resolution
Protocol (ARP) cache poisoning do not work at all. Protection is provided by
Amazon EC2 against customer who are attempting (maliciously) to view other’s
data, sensitive traffic should be encrypted by customers as a standard practice
[46].
Sharing computing resources (N9)

Sharing technology is a top threat to cloud computing because it inherits all
the problems that are possible with sharing resources in stand alone systems
[5][17][110][71]. In the world of cloud computing information is stored in data
servers which are globally distributed. This cloud computing architecture is sup-
ported by virtual machines that run on hypervisors [90]. Due to this the user
will lose control of physically securing data and this might result in security risks
because this data is stored in a location where resources (storage, computational
resources, etc.,) are shared with some other organizations [119][34]. Sharing re-
sources between different projects and products and remote storage and process-
ing of data can be beneficial but there are also some risks (such as how data is
handled and misuse in command) and can complicate computation (i.e., moni-
toring, analysis and reporting for company needs) [50][169][123]. In addition to
the complications, there are various challenges that author mentions need imme-
diate attention when sharing computing resources, such as for tenant-isolation,
customer-specific application customization etc., that can effect customers if not
using proper measures [50]. The issues that arise with shared technologies effect
both public and private cloud [142].
Table 4.3: Cloud computing challenges related to network

Network security N1
Application vulnerabilities N2
Host and network intrusion N3
Denial of service N4
Network related Men in middle of attack N5
Ip spoofing N6
Port scanning N7
Packet sniffing N8
Sharing computing resources N9
4.2.3 Virtualization
Vulnerability in Virtualization (V1)
Information from articles that discuss about vulnerability in virtualization and
cross-vm information leakage are considered.
The most commonly used multiple way to create multiple virtual machines
on a single physical machine is done using Virtual Machine Monitor (VMM)
approach. Hypervisors are also used to manage multiple VMs and any flaws in the
hypervisors can allow attacker to gain access in an inappropriate way, even when
tools such as Xen access are used many security risks can be found, which allow
admin to see through the user level process while the customer is running his VM
and attacker can easily install a malicious code [90][38][105]. In article [144][50]
the author mentions that major risk is to ensure different instances running on the
same physical machine are isolated and this requirement is not yet met completely.
If virtual machines have vulnerabilities then sharing such hardware can result in
many vulnerabilities [15]. To control and administer host, guest operating system
is another issue which needs attention. The author also argues that most of the
current existing systems do not satisfy perfect isolation and many bugs found here
are from Virtual Machine Monitors (VMM’s). These bugs can be exploited by
malicious local users and bypass some security restrictions or gain privileges. Ex.,
as been explained by the author “vulnerability in Xen caused due to an input
validation error in tools/pygrub/scr/GrubConf.py”, which can be exploited by
‘root’ users of a guest domain to execute arbitrary commands in a domain using
specially designed entities in grub.conf (when guest is booted) [144][84][44][82].
If an unauthorized user hacks into CC, he can access all the virtual machines
running in CC and control them [166]. Since CC is relatively cheap and provides
resources to users easily and also since virtual machines (which is a main part of
CC) are easy to copy or clone, the VMs can be easily used to launch new attacks
[109][67].
The following are some of the vulnerabilities by different authors mention in

their work:
• Author for [134] mentions that trusted hypervisor is more important than
having a trusted VM, because an untrusted hypervisor can affect to a greater
extent.
• The author in [123] expresses that increased usage of virtualization is a

threat.
• In [50] the author mentions insecure areas in virtual machine which need to
be secured and they are version management, customer specific application
customization, migration between service providers, etc.
• An attacker can also cause damage to a cloud that is connected to a virtual

machine by installing a bug in the source machine [126].
The above discussion makes clear that, if VM are not constantly monitored
there can be service breakage [136].
Virtual machine protection (V2)

Information from articles that discuss about virtual machine protection and se-
curing virtual machine boundaries are considered.
Multiple VM’s can be instantiated or halted in a single server (machine) to
satisfy list of services accepted. These services can also run multiple applications
which are based on different operating system environments [25]. In relation to
this author in [6] expressed his concern towards securing the boundaries of virtual
machines. Since the virtual machines created in cloud server have virtual bound-
aries (unlike general isolation where multiple hard disk drives), it’s the respon-
sibility of CSP to ensure VM’s that use common resources on the same physical
server (i.e., CPU, memory, I/O, NIC and others) are separated [6]. Also the host
needs to be secure, because if the host is not secure protecting offline/online VMs
is going to become a challenging task [38].
Virtualization in the cloud: The author expressed views on virtualization
has been vulnerable to breaches, be it on traditional infrastructure or a cloud
architecture. This virtualization is a process of creating multiple Virtual Machines
(VM’s), on a physical server to use a shared pool of resources, which also raises
a concern on possibilities of spreading malicious activity. To prevent this kind of
malicious activity in physical environment segmentation can be used, but in case
of virtualized environment there is no possibility of segmenting. But without
segmentation cloud could possibly allow hacker in web systems jump over to
financial systems or databases. The solution to such kind of problem in virtual
environments is to use Virtual Security Gateway, this enables users to apply some
critical rules, log-in and access privileges, similar to what is done with systems
placed in-house [3]. Supporting this author in [74] mentions that possibility that
dynamically assigned virtual spaces in CC can already have some security threats.
Networking in virtual networks (V3)

Information from articles that discuss about networking in virtual networks, net-
working problems, virtual management, virtual network security and hypervisor
security are considered.
Virtual machine instances interconnectivity (i.e., communication) is a huge
concern in CC [75][38]. Traditional mechanisms such as VLAN’s (Virtual Local
Area Network) and firewalls are proving to be inefficient when used in a virtualized
environment [73]. The security of a computer depends on the quality of underlying
security mechanisms or kernels which control the execution of process [78]. Also
when the devices are virtualized user might lose the visibility of how the VM’s are
connected and insecure information transfer can be possible (information of one
customer being disclosed to another customer) [163][138]. In addition to losing
visibility, when considering security at the network level in virtual machines there
arise two issues as discussed in article [6]. One of them is virtual network security,
when sharing network infrastructure between different users within same server or
physical network can increase possibility to exploit various vulnerabilities (DNS
servers, DHCP, IP protocol, etc.) [6]. The other issue mentioned by [6], is
on hypervisor security, this is the ‘virtualizer’ that links physical resources to
virtualized resources and virtual machines in a virtual environment are given
access to physical resources by this hypervisor. Hypervisor provides isolation
between the different guest and virtual machines. Since these hypervisors are
virtual machines that run on physical machines and also smaller in comparison
to physical systems, this helps in controlling and isolating processes. But now
a days hypervisors or virtual machines are large, complex and comparable to an
operating system [78]. Hence understanding hypervisor security is critical, any
compromise on these machines can effect security of all users connected through it
[6]. Examples of hypervisor software’s are VMware or Xen. Also the author in [56]
mentions that the auto arrangement of virtual resources can lead to problems and
suggests that SPs need to perform security checks and constantly verify if user’s
security requirements are met or not. One way to see that VMs are protected and
not controlled by malicious users, administrators since they are the key persons
in managing VM, the CSP should regulate the administrators from time to time
[166].
Within most of virtualization platform existing there lies inbuilt ability, to
create software based swatches and network configurations as part of the virtual
environment and provide communication between virtual machines which is more
efficient and direct. Traffic over virtual networks may not be visible to security
protection on physical networks, so there is a necessity to duplicate such services
in virtual networks [78].
Some common attacks that can effect virtual machine’s network are “buffer
overflow, DDoS, zero day attacks, viruses, covert channels, trojans, etc” men-
tioned, which can infect the hypervisor as it is the major VM controlling entity
[5].
VM isolation (V4)
Inter positioning is an inherent feature of VMM, isolation of VM is a key concern
as it is not that possible CC and active inspection mechanisms are yet to be
designed [73][75]. Multi-tenancy is a process where multiple cloud customers have
their virtual instances running on the same physical server. In such environment,
there might be a possibility for malicious user allocating an instance in same the
server and possibility is that he/she (malicious user) might penetrate the isolation
between the VM’s and exploit them [128]. This could be possible by creating a
side channel between two VM’s on a same physical machine and enabling a SSH
keystroke timing outlined attack [34][141]. There is also a possibility that data
accidentally crosses the virtual boundary [136]. The author in [126] mentions
that even in an isolated environment if an attacker interferes with the VM placing
strategy and position their instance on a physical machine of the victim, all the
private information of target was visible.
To secure VMs, the VM instances need to be isolated, but at what level
isolation should be implemented i.e., is it at VM, physical machine, local area
network, VMM or at data centers [73]. The author in [105] mentions that logical
separation need to be validated. Will that be sufficient? But there are more
problems to handle.
Author in [38] mentions about protection from the interconnection of the VM,
which connections are going into and which are coming out. The author explains
various issues that can arise with existing technologies. The author [38] also
mentioned there could be issues arising in the following areas of VM if isolation
of VmM’s is not done properly by:
1. Monitoring VMs from the host
2. Communication between VM and host
3. Monitoring Vm’s from other VM’s and virtual machines in mobility.
VM Securitization (V5)
Information from articles that discuss about VM securitization and VM security
are considered.
VM introspection or behavioral monitoring in virtual machines requires a lot
of computational power. It is easy to clone and distribute VM’s between physical
servers but this distribution could result in propagation of configuration errors.
Maintaining a hypervisor level, Media Access Control (MAC) and trusted com-
puting techniques are suggested mechanisms used to build future secure cloud sys-
tems. In IaaS virtual machine security lies with the cloud consumer, a consumer
can use their own security controls based on situation or security management
process and expected risk level assessment [6].
Table 4.4: Virtual machine related challenges in cloud computing

Vulnerability in virtualization V1
Virtual machine protection V2
Virtualization Related Networking in virtual networks V3
VM isolation V4
VM securitization V5
4.2.4 Organization
Organizational security management (O1)
When adapting to cloud computing, some changes are introduced to the secu-
rity management, information security lifecycle models, even the corporate IT
standards and policies need to be changed [160]. There are issues such as less
coordination among different communities of interest within client organizations.
The customer also has to face new risks introduced by perimeter-less environ-
ment, such as data leakage due to multi-tenancy, issues like local disasters and
provider’s economic instability. But since the cloud computing environment is
distributed in nature, re-evaluate best practices and adoption of secure cloud
computing applications become extremely complex as they require to have a well
structured cyber insurance [146]. Another way is to adjust to the new features
provided by the cloud computing else gaining complete benefit from CC would
not be possible [50]. Some point needed to be considered and as follows:
• Will proprietary interfaces used to administer services provided be accessi-

ble? [65]
• Security models must be assessed in detail [142].
• Gather and manage security related information on non native interfaces

such as CC is important [91].
• Some organization policy and legislations are non compliant with CC, ex-
ample enterprise network which uses NAT (Network Address Translation)
and company specific firewalls and is not easy to integrate with CC [97][49].
Trust management and Policy integration (O2)

Trust is the key problem in the CC environment as CC is not completely trustable
for users [164][136][67]. An organization using cloud computing application has
to give some of its critical information to use certain services provided by the
cloud service provider. For this organization needs to develop trust in remote
execution/storage and placing sensitive data in the cloud is tough as the com-
pany feels that its loosing control over data and there is much less transparency
in cloud [89][77][4][32]. Lack of this control is leading to distrust in CC [123].
The remote execution in CC is carried out based on some policies and principles.
There are evidences that even though the policies at individual domains remain
verified, the violation is still possible to occur at integration that needs to be
taken care of or have some attention. Hence, there is a need for a trust based se-
cure interoperation which helps in managing trust and support policy integration
[146]. Insider access, data stored or processed outside organizational boundaries
and security controls are some features which when used with cloud computing
could have some inherent risks. Threats that arise with the insider access are
well known for various organizations. Its influence over cloud computing has not
at all changed [78][84]. These threats could be raised by a user or a third party
organization that was given access to a part of application in the cloud. These
threats could result in theft of critical information or various types of frauds
[78]. The CSA mentions that violation in cloud can be caused by a third party
provider when it tries to access users data stored in the cloud [7]. The author
for [4] mentions that trust in customers can be increased when SP develops trust
worthy tools, automated processes for working in cloud and methods that estab-
lish healthy connections between user and provider. The author also mentions
other attributes of CC such as, compositional chain of trust, trust re-evaluation
and transparency vs trust evaluation [4]. The author of this article illustrates
multiple examples of possible threats and possible security measures [179]. Since
there are no extensible trust building models available in CC, SLAs should be
monitored and enforced properly as they are one way to build trust between user
and provider [38][64]. Mutual auditing users/providers in the CC will help in in-
creasing trust and assurance on others in the cloud [110]. In article [148] author
suggests a need for trust building approach to negotiate, establish and maintain
trust. He mentions that this should adaptively support policy integration. The
author poses some questions that need to be answered [148]:
1. How to manage and maintain dynamically changing trust values
2. How to adapt to the access requests as trust keeps evolving
3. The most important one, how to establish trust and determine assess map-
ping which satisfies inter-domain access requirements
Failure in providing security (O3)

Information from different articles that discuss about fault tolerance and failure
in providing security are considered. Failure in providing security to the infras-
tructure under control of cloud provider can result in compromising subscriber’s
security. Even a single weak link in cloud computing can cause a security threat to
multiple entities connected in it. For customer to secure his data, he/she should
believe in service provider’s security [154][87][46]. It is evident that customer must
also have trust in providers security. For a cloud to be trusted and considered
strong simple but yet important features such as logging, security policies,incident
response, etc., should also be strong [90]. A third party poor management can
lead to damaging the provider and also the user indirectly [138].
In addition to the above discussion, article [127] gives us a few examples that
are tabulated in table 4.5, which exemplify different cloud failures.
Table 4.5: Examples of cloud computing previous failure
Service and outage Duration Date

Microsoft Azure; Malfunction in windows 22 hours March 13-14, 2008
Azure
Gmail and google apps engine 2.5 hours Feb 24,2008
Google search outage:program error 40 minutes Jan 31,2008
Google site unavailable due to outage in con- 1.5 hours Aug 11,2008
tacts system
Google AppEngine partial out- 5 hours Jun 17,2008
age:programming error
S3 outage: authentication service leading to 2 hours Feb 15,2008
unavailability
S3 outage: Single bit error leading to gossip 6-8 hours Jun 20,2008
protocol backup
Flexi Scale core network failure 18 hours Oct 31,2008
Identity and access management (O4)

Information from articles that discuss about identity management system, iden-
tification, identity and access management, identity provisioning (Independent
IDM stack, credential synchronization and federated IDM) are considered.
Identity management is an administrative process area, where all users of
a system are identified (entities), and by enforcing some restrictions on these
entities access to resources in that system are controlled [144]. Since this plays a
key role in securing Cloud Computing (CC) environment access should be given
only to privileged users, which should also be blackened by physical monitoring
and background checking [92]. The major concern in this is how complex it
is to manage a diverse population of users, and provide access to internal and

external services in a constantly changing environment (changes can happen due
to change in business needs/processes) [104]. It is also true that need for fine
grained access control on the information stored on CC is increasing and since
CC is vast satisfying these needs and meet growing requirements has become
complicated [170]. For this there are various mechanisms but not all are up to
the satisfaction of users [170]. The browser based authentication protocols are
not secure as they cannot issue XML based security tokens [75]. In addition to
this, organizations may contain complex web of user identities, access rights and
procedures.
Author of [126][105] says access control in CC is an issue and it is necessary
that a user is able to view only the portion of data he is given access. Access
control models should also be able to a pick out relevant areas of SLA and change
access rights accordingly [148]. Generally organizations give access to its employ-
ees based on principle of least access, where access is given to only those services
to which he/ she needs basic access [119]. Identity is the core of any system and
to have a manageable control on users and identify breaches, there are various
areas need to be considered: “Identity Provisioning and deprovisioning, identity
information privacy, identity linking, identity mapping, identity federation, iden-
tity attributes federation, single sign on, authentication and authorization” [6].
Identity management can be done in three different ways [144]:
• To just create, manage and delete entities without regard to access or knowl-
edge regarding the access rights of the entity. This is known as a pure
identity paradigm.
• Access can be provided to users with the help of some smart cards or a
traditional scheme of authenticating. This enables user to log on to a system
for accessing a service or services, which is other way, called as the user
access (log-on) paradigm.
• Providing access to users on various devices based on different distinctions
such as Role based, online, on-demand, multimedia content and presence
based, is termed as a service based paradigm.
A traditional way of having control over an application was to keep track of col-
lection of users and manage them and they are still not sufficient in CC [170][112].
This system is no longer easy as sharing of Personal Identifiable Information (PII)
is not suggested because doing so can cause a breach to privacy. The PII is the
information that a user sends to Service Provider (SP) to authenticate it. PII dif-
fers from user to user [124][18]. Some suggested methods for managing identities
are [144]:
• Independent IDM stack: In this system, all information related to user
accounts, passwords, etc., is stored with the Cloud Service Provider (CSP)
and it provides management and sign on services.
• Credential synchronization: In this system, user account information is

replicated between enterprise and service provider. The user account infor-
mation is done at each guest who pays for the services provided by the CSP.
This replication is done at each user based on their company regulations.
Replication also considers relevant portions of user credentials at the CSP
to provide sign on and other access control features for services used in CC.
• Federated IDM: Within this kind of system, user credentials are stored
and managed at the enterprise itself. Authentication is done in enterprise
boundary itself and if necessary, certain attributes are communicated to the
CSP [144][165].
Another issue here is that companies having a strong IDM system might not
be able to extend its functionalities on CC or change the existing framework to
support this new architecture of cloud [148]. To separately maintain a different
authentication for internal (organization authentication) and external (authenti-
cation at SP’s end) is a complex solution to this problem [78]. To nullify the effect
of these challenges on CC, systems need to find possible breaching zones/areas
before they cause some impact. Given below are some assessment test to validate
the security in IDM [144]:
• Authentication weakness analysis
• Insecure trust configuration
The author in [90] explains a special case, by taking an example of a employee

with malicious intent, working in a company which provides CC services. This
employee can easily gain access to secure areas and perform malicious activities
which could be difficult for outsiders. The impact of this employee could be se-
vere if this employee is not identified by the service provider. Supporting the
previous example, article [142] mentions that more than 67% of respondents in a
survey by CERT insider threat center 2010 cyber security watch survey admitted
that insider incidents are the costliest. The author also suggests to use a in-
sider security policy to avoid insider attacks [142]. The author for [166] expresses
that insider attacks are alarming and this might happen when security standards
provided by SP do not meet the standards used by customer’s standards. The
author of [109] mentions that one third of breaches happen form lost or stolen
devices and from employees sharing their data accidentally on the internet and
nearly 16% due to insider threat. This issue of malicious insider has also been
mentioned in following articles [17][5][105]. Not only there is an insider threat
but even a malicious user might gain access to CC and deploy malware if he can
produce a stolen credit card from legitimate user [138]. These insider threats or
entry for a hacker or malicious insider is easy because SPs don‘t give customers
transparency into security in CC. Inefficient access and identity management can
lead to illegal service process, fabricated services process and malicious service in-
terruption [32]. The author for [43] also mentions various types of insider and the
impact and also mentions that it can be dangerous if go unattended. Article [91]
mentions Identity management, credential management, attribute management,
privilege management, cryptographic key management as security management
areas in CC.
Authentication and authorization: Most companies store their employee in-
formation in some kind of Lightweight Directory Access Protocol (LDAP), Small
and Medium Business (SMB) companies which have SaaS as their highest adop-
tion rate seem to use ‘Active Directory’ as the most popular tool for managing
users. In case of SaaS, the user credentials get stored multiple times in provider’s
databases, because SaaS requires the user data to be stored in SP’s data repos-
itories. To avoid misuse SaaS customers should remember to remove/disable
accounts or create/enable as employees leave or join the company [144].
Authentication and identity management: Identity management and authen-
tication is being realized as a critical requirement in CC [66][163]. Users can be
authenticated based on service based credentials and characteristics using iden-
tity management (IDM) mechanism. This system is supposed to protect private
and sensitive information related to users and processes [162]. In multi-tenant
cloud environments customer’s identity and authentication information should
be segregated. These components need to be easily integrated table with other
security component. It is not clear yet on how multi-tenant cloud environments
can effect privacy of information [146].
Most of the service providers use Security Assessment Mark-up Language
(SAML) in administering and authenticating user before giving access to re-
sources. This language provides an interface where information related assert-
ing subjects or authentication information is exchanged between cooperating do-
mains. SAML request and response messages are mapped over Simple Object
Access Protocol (SOAP) which in turn depends on extensible Mark-up Language
(XML) for formatting. This SOAP message security validation can become com-
plicated and threat prone. As this SAML depends on XML, there are some
evidences where wrapping attacks on XML were successful and in turn that can
be effecting SOAP message. This in turn can become a threat to the SAML
authentication procedure [78]. The detailed description on challenges with im-
plementation of other modes of authentication such as, single sign-on, OpenID,
RBAC, etc., are given in [66].
Given below are some cloud specific issues mentioned in [62] article:
• Denial of service by account lockout: The most general way of authenti-

cating a user is by username and password. To avoid misusing an account,
after a certain numbers of failed attempts an account is locked. This fea-
ture can be used by an attacker to launch a Denial of Service (DoS) attack
against an authorized user [144][38][20][143].
• Weak credential-rest mechanism: Password recovery mechanism prevailing

is proven to be weak, if a Cloud computing (CC) provider manages user
credentials manually than using a federated authentication, there should
be a method provided for resetting credentials in case of lost or forgotten
credentials.
• Insufficient or faulty authorization checks: Web applications and service

cloud are vulnerable to faulty or insufficient techniques of verification that
can make users access unauthorized information. URL guessing attacks
are the outcome of missing some authorization checks, the user can modify
URL’s to display other user’s information.
• Coarse authorization control: Cloud service management interface needs

to offer the authorization control models, which are still not refined. This
makes standard security mechanism (duty separation) not so easy to be
implemented in a cloud environment.
• Insufficient logging and monitoring: Present day cloud has no standards or

mechanisms that can help customers to monitor and log data within cloud
resources. This can cause problems when the log file needs to be searched
for a single tenant and also there is not all possibilities from CSP’s side to
monitor user’s data.
Authorization and authentication is an important information security re-

quirement on which cloud architecture designer must concentrate and build a
secure and trusted cloud computing infrastructure [123][8]. Inefficient authoriza-
tion mechanism can cause side-channel attacks in the cloud [8]. The author in
[171] gives an example of a special case where two parties using the cloud as a
medium for transferring applications/code/information, only the target user is
supposed to get access to the piece of program but with the help of an insider
(an employer in cloud) the target user can create another copy of the same code
and sell it. This is a copyright issue which can cost the developing company.
Traditional encryption, hashes or access-control labels can not help to avoid this.
Which gives an example that shows if access management is not given enough
importance a user who is not fit for having access to a part of the transaction can
misuse his rights and cause damage to customer. To avoid this proper monitoring
should be done for employees in cloud.
Some suggestions to prevent illegitimate access are as follows:
• Mentioning human resource requirements as part of SLAs, transparency

in overall information security and management practices, compliance, re-
porting and determining a security breach notification method and the most
important enforcing strict supply chain management are some suggestions
given by CSA [90].
• Before committing to a SP the customer must make a note about the privi-
leged user access policies, know who will be gaining access to the data, how
can it be accessed etc., [55][85][123].
• An effective and continuous monitoring system which detects malicious ac-

tivity will be better than depending on firewall and security mechanism
[105].
Availability (O5)
Information from articles that discuss about availability, cloud provider goes
down, job starvation due to virtualization, business continuity and service avail-
ability are considered.
Availability is a primary concern for CC and is one of the key issues in in-
formation assurance [169][65][166][164]. Not being able to access services can
become a serious issue [169]. It is evident in literature that the availability of
a system is important and many organizations/researchers have their questions
such as “if downtime occurs can the business still operate?” [162].
Availability of a system can be defined as, the extent to which an organization
provides its user with round the clock service (resources accessible and usable)
is availability. There are chances of having a complete or permanent impact on
the organizations when the availability of a system is affected. Denial of Service
(DoS) attacks, natural disasters and equipment outages are all threats to avail-
ability [78][134][31][100]. The cloud services also have experience in outages and
performance slowdowns; if not planned and designed properly customers using
cloud computing services will be at loss. There are various examples illustrated
by [78], where the effect of shortcomings were borne by well-established organi-
zations such as Twitter (in 2009), lighting storm caused some IaaS customers go
down (in July 2009), gmail (in 2008 for one day) etc., [78][35][7], these outages
can also occur because of non-technical issues [13][12]. Service interruptions could
also be due to malicious users/use [32]. If at all there is an outage or server crash
all the VM, services running on that server become unresponsive, in addition to
that high availability or automatic fail over or high availability is not possible
in public cloud [138]. If an organization wishes to migrate its servers which is
so vast that it has multiple data centers distributed over different geographical
locations using a variety of network providers but with some common software
service to connect with a cloud, in such situation the organization might not want
to migrate to a new system without a good business continuity plan modeled. In
this case, it is suggested to have a multiple service providers providing differ-
ent stacks, such that if one stack fails there will be another stack (i.e., service
provider) supporting as a substitute [13][12][72][116].
Generally CSP has to provide required application level scalability and avail-
ability, by making necessary changes to the applications at the architecture level
and satisfy the customer requests. CSP can design specialized hardware and soft-
ware to deliver higher reliability and earn an additional penny out of this [13][12].
CC should be constantly available with all the access control mechanisms and
data access working [70]. To make this possible, a multi-tier architecture is de-
ployed with a load balancing architecture running on a variable number of servers.
To enhance the availability of a system built in CC, it is also suggested to secure
system for hardware/software failures and DoS attacks. For safeguarding data
from unplanned emergencies or minimizing enterprise downtime there is a need
to have an appropriate plan for Business Continuity (BC) and Disaster Recovery
(DR). For example the amazon API’s use a Standard Distributed Denial of Ser-
vice (DDoS) mitigation techniques such as synchronous cookies and connection
limiting. Amazon also maintains a higher level internal bandwidth than supplied
to its customer, to further ensure mitigation of DDoS attacks. To validate the
vendor’s availability, given below are few tests [144]:
• Authentication weakness and
• Session management weakness
Another solution for maintaining availability is, securing the system by usage
of a Trusted Third Party (TTP). TTP provides end to end security services
which are scalable across different domains and usable with different geographical
areas. TTP’s are impartial organizations delivering business confidence through
some technical security features and securing an electronic transaction. These
TTP’s are generally connected through a chain of trust to provide a web of trust
using the Public Key Infrastructure (PKI). PKI in a distributed information
system becomes effective when combined with a directory (it is a set of similar
attributes organized in a hierarchical and logical way). The Light weight Access
Protocol is internet’s (LDAP) general way of accessing a directory service. In CC
environments PKI used in combination with Single Sign On (SSO) mechanism
is ideal as they provide smooth and transparent authentication across various
physical resources. The reasons for depending on TTP’s are as follows [179]:
• Low and High level confidentiality.
• Server and Client Authentication.
• Creation of Security Domains.
• Cryptographic Separation of Data.
• Certificate-Based Authorization
Usage of TTP or any other method used for ensuring availability is generally
mentioned through the Service Level Agreements (SLA). To ensure greater extent
of availability it is important to identify the right deployment and delivery model
set out by the International Organization for Standardization (ISO) [123]. Usage
of HTTPS+WS security can cause heavy load on the server leading to service
break down (reasons for outages in google and amazon in 2008), there should be
perfect balance between availability and security measures to keep services active
[123].
Table 4.6: Organizational security concerns that challenge cloud computing

Organizational security management O1
Trust O2
Organization related Failure in providing security O3
Identity and access management O4
Availability O5
4.2.5 Scalability
Incompatibility issue (SC1)
Information from articles that discuss about portability, interoperability, scaling
quickly, smart scaling, automatic scalability, security management function, col-
laboration, integrity of security, general requirement cloud security management
model are considered.
For instance there might be a situation arising where a user wishes to change
his Cloud Service Provider (CSP), in such situation the new CSP should be
capable in moving data from an old service provider. This transfer includes data,
components and also security policies. In most cases user won’t be provisioned to
take physical firewall but users should make sure to have a copy of its configuration
on virtual machines [3]. There are chances of customer goes locked to a SP
if the customer chooses a wrong service provider [109][105]. Cross data center
operations would find this as a serious issue [55]. In article [91] author mentions
that controlling defining and ordering in relation to security are one of the ten
things to be noted while shifting to cloud computing.
CSPs are creating the hosting world using “sticky services”, these services
are reasons for causing difficulty in moving between different service providers.
Example, the Amazon’s simple storage service S3 is incompatible with IBM’s
blue cloud or Google or dell [24][130][46]. Open cloud manifesto which was newly
published was declined by both Microsoft and Amazon and are pursuing interop-
erability on their own terms [46][16][24][173].
Things to be noted before/after moving from one CSP to another are [3]:
• In infrastructural service, backing up of data is easy compared to other

services.
• In case of web application, there should be a perfect plan and process on

how and when to move data. If all of your data is visible then it is not
necessary to move it completely, but instead necessary data can be selected
and copied.
• Application migration and reconsidering application infrastructure are ar-
eas to be considered by infrastructure architects. This is because some
companies such as Microsoft are provider-specific and are not built to be
easily reused [50].
When data is stored in the cloud, there are some virtual perimeters and secu-
rity models with some shared responsibilities, among CSP and customer. These
shared responsibility models will induce some changes into organizations IT staff.
To maintain stability in the organization, the Chief Information Security Offi-
cer (CISO) is required to know if the cloud service provider allows its user to
control and manage implementation of security policies and get assurance that
business data in cloud stays protected. To assure data stays protected there are
two considerations, one is that there should be an analysis made to find out what
additional security controls need to be implemented on existing controls provided
by CSP. The other point to be noted is how enterprise security management
tools and process adaptation manage security in cloud [104]. Policy integration
task in the cloud should also be able to address secure interoperability, semantic
heterogeneity and policy evolution management [148].
Scalability of storage: One advantage with cloud computing is that it can
provide infinite capacity based on demand of the user. Its not yet clear on how
persistent storage systems could be when it is practically implemented. The at-
tempts to provide these scalable conditions was done by trying to vary richness of
query and storage API’s. There is still a challenge in this area for the researchers
i.e., not only to meet the expectations of programmers in providing durability,
high availability and ability to manage data quickly but also to combine these
qualities with cloud where advantages of scaling up and down can be used effi-
ciently [13][12].
Interoperability: Interoperability is an important to maintain global stabi-
lization to reduce rework and management [41]. Organizations willing to adopt
cloud computing wish to have an Identity Management (IdM) solutions that
can interoperate with all the existing IT systems and solutions, with or without
small changes. Some of the most commonly used authentication mechanisms are
Microsoft Windows authentication, SSO, LDAP, SAML, OPENID and OAUTH,
OpenSocial, FaceBookConnect, etc., and cloud users expect easy integration with
these types of services. To conclude CSP would have to provide an authentication
module where each and every type of authentication system that user wishes, to
use are easily comparable with cloud system [61]. Organizations are interested in
CSP’s, who provide a framework which can enable interoperability with different
other Service Providers (SP), Amazon API’s standards, are the most commonly
used standards to design other on demand instances. CSP’s which use these
API standards are easily comparable with Amazon (Amazon EC2 can run eas-
ily in Eucalyptus). In certain situations cloud service providers which assure
to provide interoperability with other SP is also problematic, example Hadoop
which provides on demand capacity can’t for sure provide compatibility to run
on another system developed using C++. Several organizations such as Cloud
Computing Interoperability Forum and Open Cloud Consortium are still try-
ing to provide firm and stable standards for CC. There is a framework named as
‘Thrift’ which relies on code-generation engine, to provide scalable cross-language
services development. To provide interoperability and compatibility with various
programs there could be a common language, which can be used by different
service providers. As a solution to this several people attempted but there is
no single language that is up to the requirement [63]. A panel report on cloud
computing standards discussed about proposing “open standards and Predictive
Markup Language (PMML)” and expressed views that implementing these could
be extremely helpful not only for software vendors and data mining community
in general [172].
Constant feature addition (SC2)

In cloud computing scenario users need to be updated with new security features.
These updates need to be done by the service provider and are required to be
quick. The rate at which updates are installed can effect security as well as
software development lifecycle [46].
Scalability (SC3)
CC requires the ability to scale multiple transactions managed over multiple con-
nections. Scaling the resources manually is still persistent with IaaS, where SP
will define how the service has to be scaled based on personal experience and
different factors influencing it [109]. There should be a reduction in management
tasks and automatic scaling with-in the CC applications should be enhanced
[131]. With software today it takes approximately 6 months for a single SAM-
L/SSO connection, which does not address the compliance and access control
issues [151]. Open cloud manifesto states that, to keep out of problems with al-
location of resources during peak hours cloud services have to dynamically scale
up and down [61].
Table 4.7: Scalability related concerns which challenge cloud computing

Incompatibility issue SC1
Scalability related Constant feature addition SC2
Scalability SC3
4.2.6 Confidentiality and Privacy

Confidentially (C1)
Confidentiality plays a major role in securing organizational content stored across
different databases [65]. It’s a key issue and since most data is virtually accessed,
protecting and maintaining confidentiality of user profiles is of utmost considera-
tion [123][136][177]. Viruses, trojans, malwares, etc., are some unauthorized ways
to exploit customer’s information [67]. In some cases it is also important for an
organization to handle data remains, this is to protect the confidentiality of an
employees information even after his data is removed or erased. Remanence is an
issue that can lead to the disclosure of private data [179]. In article [48] author
mentions that even though there are methods to provide confidentiality they are
not widely used by SPs. As a special case in [142] author mentions that because
of confidentiality agreements made with user, preventing/identifying malicious
attacks is becoming difficult (top threat mentioned by CSA). This is because a
SP can not monitor or look what’s happening in the user’s space due to confiden-
tiality agreements, which can be exploited by malicious users for unauthorized
activity [142][90].
To prevent data being disclosed or misused in cloud and protect customer’s
data, every customer should be educated as to how data is stored in the cloud.
By doing this the user will be careful while storing sensitive data in the cloud
[160].
Privacy (C2)
Privacy is one of the cloud computing security requirement [92][65][74][109][160][54][126].
Keeping data private in a distributed system is challenging when compared to per-
sonal possession and in CC it is risky [149][23]. Privacy or obligation is related
to the collection, use, storage, disclosure and destruction of data that is personal
to someone. The rules and the concept of privacy varies with countries, cultures
and jurisdictions [104][111][74]. The author in [179] mentions privacy as a desire
to control disclosure to his personal information and presents that there are a
number of legal challenges to cloud. Privacy is being accountable to an orga-
nization’s data subjects and also be transparent towards organizations practice
around personal information, there is also a little knowledge on how privacy laws
that govern within an organization [111]. There is no universal agreement to-
wards defining what constitutes personal data. “The rights and obligations of
individuals and organizations with respect to the collection, use, retention, and
disclosure of personal information.”, is one way of defining privacy and this is
gaining popularity among American Institute of Certified Public Accountants
(AICPA) and the Canadian Institute Of Chartered Accountants (CICA) in the
Generally Accepted Privacy Principles (GAPP) [104]. As a special case in [142]
author mentions that because of privacy agreements made by the user, prevent-
ing/identifying malicious attacks is becoming difficult (top threat mentioned by

CSA). This is because a SP can not monitor or look whats happening in the user’s
space due to privacy agreements, which can be exploited by malicious user for
unauthorized activity [142][90]. And this case preserving privacy i.e., protecting
the private data from unauthorized users in cloud computing while maintaining
sharing of resources is a security issue [177][17][140][75].
Services can keep varying among customers and also the service providers
in CC, since these services keep changing private data and personal data moves
within an organization or could also cross organizational boundaries and pro-
tecting such information is important. Fields such as financial and health, are
concerned about safety of data. Following are most important privacy risks, which
need to be covered [117]:
• In case of a user, he might be forced or persuaded to give his personal
information against his will [49].
• From organizations perspective, compliance with origination policies, legis-

lation, creditability and loss of reputation are some other issue.
• With implementers of cloud platform, possibility of exposure of sensitive

information that will be stored on platform, legal liability, loss of reputation
and credibility, lack of user trust and take-up.
• For providers of applications designed over the cloud platforms, loss of rep-
utation, legal noncompliance ‘function creep’ which uses the personal infor-
mation stored on the cloud (i.e., it might later use for purpose other than
that for which it was originally intended).
• In case of data privacy, exposure of personal information. To prevent loss

of personal information there is a need for a special committee, which keeps
track and makes decisions related to data privacy. If a security compliance
team already exists within an organization it won’t be having formalized
training on data privacy, a possible solution to this is to hire a privacy
expert or train an already existing member well [119].
• Organizations can be held liable even if the subcontractor causes security

breaches and CSP is legally considered same as that of a subcontractor.
This confirms that organizations should ensure that a CSP is/are compli-
ant to respective privacy legislations. Various governments have posed laws,
which make them accessible to data stored in their jurisdiction for electronic
discovery or anti-terrorism purposes. To gain access to data stored in the
cloud i.e., search for data in cloud government should in most cases issue a
search warrant and this can differ from service provider to a service provider.
In some cases CSP fails to provide required computing resources, in such
situations CSP may be forced to outsource data for a different CSP. The
subcontracting CSP’s don’t inquire about compliance with privacy regula-

tions when establishing a relationship. The organizations are not aware of
these privacy regulations or they think these are not important to comply
with. Raising awareness about both issues and existing regulations seems
like a good step to solve this issue [132].
The author of [166] suggests that privacy of user data and laws regulation and
policies need to be framed taking CC into context, which will prevent security
and user involvement in using the CC application. Methods like homomorphic
encryption technique, which helps to protect data will complicate handling huge
amounts of data and in an environment where data keeps growing is tough [23].
Providing individual user control, anonymous services for individual use, limiting
identity information, and requiring authentication for high level transactions are
some of the features that safeguard privacy of users. For this SPs have to encrypt
the user information, isolated data processing and storage, managing privacy
and security requirements are some of the issues to be dealt while working in
cloud computing [122]. Also transaction histories, identity information, policy
components during integration, etc., need to be included and protected [146].
Table 4.8: Challenges faced by cloud computing in relation to confidentiality and

privacy

Confidentiality C1
Confidentiality & privacy
Privacy C2
4.2.7 Backup and Recovery issues

Backup (B1)
CC servers are place where users store all the sensitive enterprise data and regular
backup of the user data needs to be done as a fault tolerant mechanism and
recover case of disasters where original data is destroyed [144][166]. But the
author of [105] is concerned what will happen to the data backup if the company
switches? Or company goes down? He also mentions relying on CSPs backup
could be foolish. There is also another concern from customer point of view,
which mentions that will data stored in the cloud will still be valid even though
the cloud provider go broke? Will the data stay intact, accessible, without any
logistical problem even when there are merged and acquisitions made by the
service provider (long-term viability) [85][142][123][109][55]. The main aspect in
all this discussion is to verify whether the client data has high probability in server
side. Malicious vendors try to make it fake and collect the data from the server.
For example: the server claims that it is storing five copies of data but actually
it is storing three copies of data and shows only 5 copies of data occupation [42].
1. Insecure storage
2. Insecure organization
For this there are assumptions such as:
1. Trusted Platform Module (TPM) is installed for each data backup.
2. Private key is certified by the third party.
3. Assume that the server cannot launch any sophisticated hardware to the
stored data.
Article [116] exemplifies a situation of issue when a backup is not properly man-
aged.
Data retention and recovery (B2)

Disaster recovery is another important issue [55]. To recover data service provider
needs to have business continuity and disaster recovery planning policies [142].
Even if the customer do not know where his/her data is, cloud provider should
be able to tell what will happen to it in the event of a disaster and how long
will it take to recover? [85][123]. Industry pundits warn that if any offering is
made which does not replicate the data and application of infrastructure across
multiple sites is ‘vulnerable to total failure’. Data replication policies should be
established along with the proof that the vendor can enact a complete restoration
and indicate them how long will it take [22][72]. The author from [164] mentions
that disaster and recovery are paid more attention in PaaS.
Investigative support: Investigating data in certain cases is necessary, and
data stored in cloud provides some complexities [55]. Author in [71] mentions
that digital forensic investigation of information, which include seizing system
for investigation in cloud is complicated. In a report on security issues of cloud
computing, Gartner pointed out that investigating inappropriate or illegal activity
in the cloud is impossible because data of multiple users could be co-located or
could be spread out in an ever changing set of hosts and data centers, which was
also agreed in the articles [86][3][142]. It also says that the only way to safeguard
your data on cloud is to ask the provider for any previous evidences of any such
investigation supported i.e., ask provider if it does have ability to investigate any
inappropriate or illegal activity? [123][142]. The authors in this context have no
evidence of any such investigations successfully performed, which means there is
no investigation possible when cloud services are used [3].
Risk management: Risk management is a process of identifying and assessing
risks and plan accordingly to mitigate or the reduce impact of risk. In CC ser-
vices some components, subsystems or complete system could be distributed and
may not be under the control of the organization using them. Most likely orga-
nizations have a better risk management when there is control over process and
equipment. With traditional information systems, risks are managed through
systems lifecycle and in case of CC services, assessing and managing becomes
challenging. Since the organization does not get full control, organization should
ensure that security controls are implemented correctly and they are operated as
expected. Organization’s trust towards cloud service varies based on the extent
of control provided to the organization for its data, applications and also on ev-
idence provided about the effectiveness of those controls. Performing all these
tasks to estimate the functionality of service provider is difficult, so solutions to
this is using third party audits and establish trusts based result. Finally if the
level of trust falls below the level of expectation and no compensatory controls
can be employed then the organization has to reject the service or accept with a
greater risk.
There is utmost need to have control over physical host security and also
virtual machine. If the physical security is compromised all virtual machines
residing on that specific host get compromised [78].
Table 4.9: Data backup and recovery related concerns cloud computing

Backup B1
Retention or backup
Data retention or Recovery B2
4.2.8 Other concerns

Auditability (A)
Information about auditability, audit and monitoring, audit and compliance, au-
ditability and data, confidentiality, compliance and auditability from different
articles are considered.
Difficulty to audit is another side effect of cloud computing which shows, as
there is lack of control in cloud [51]. Current auditing scheme in CC goes on
with the help of documentation or manual audits. Audit is an internal, external
entity that organization implements to identify requirements that organization
must abide with and helps to put into practice those policies, procedures and
process which are needed to satisfy such requirements. Audit is also used to con-
stantly keep checking if policies and principles are followed within organization or
not. For organizations to use cloud computing platforms, cloud service providers
(CSP) has to maintain, monitor and demonstrate on going compliance with cus-
tomers business and regulatory requirements. Sustainability is not possible while
maintaining separate efforts on different regulations or standards. Combination
of internal policy, regulatory compliance and external auditing should be used
[104]. When data is stored in remote locations auditing can become challenging
or cost effective when doing it on site [7][126]. Also managing auditing is given
as one of the 10 security management areas of CC in [91].
Data confidentiality/auditability: Security is the most common area for critic
by CC analysts, who keep asking “who would trust their essential data to be
stored somewhere?”. Many cloud computing security threats are similar to those
faced by data centers. This responsibility is divided among many parties such
as vendor, cloud user and third party vendors that CC users depend in securing
sensitive information. In cloud computing architecture user is responsible for
his securing at application level; cloud service provider is responsible at physical
level and enforcing firewall security. These responsibilities can also be outsourced
and given to third party service providers who sell especially security services.
Additional features such as firewall rule analysis can be provided by standardized
interfaces of platforms such as Amazon EC2. With cloud computing, internal
security threats are more when compared to external threats.
Virtualization is key ingredient in cloud computing, with many benefits and
at the same time it also brings numerous threats. Incorrect virtualization code
might allow user to access sensitive portion of information of other user or provide
access to service provider’s infrastructure. This all happens because virtualization
software contains some bugs, which might allow virtualized code go loose to some
extent. The service provider by default controls the software stack bottom layer
which effectively circumvents most of the known security techniques.
A common problem that exists not only in case of cloud computing is, data
lost into public and reason for this could be disposing a hard disk without being
wiped or a bug within the program that makes data visible to unauthorized users
partially.
Similarly, auditability could be added as an additional layer, which is kept
out of reach for virtualized guest os, this provides arguably more security than
those built into application. Such new features reinforce the cloud computing
perspective of changing focus form specific hardware to virtualized capabilities
[13][12]. In addition to all these mutual auditing should also be supported to
cross check stakeholders in CC [110].
It might be easy, or too easy to start using CC services but hard to govern
cloud related activity.
Compliance (COM)
All the compliance related terms (such as compliance, regulatory compliance,
audit and compliance) are identified from different articles are considered.
Countries have their own security, privacy and regulatory laws at different
levels (i.e., national, state and local), which makes compliance a complicated
issue for cloud computing. Compliance requires conformance with the local es-
tablished specifications, standards, regulations, or laws, which is hard for CC to
demonstrate [135][78][92]. The customer must check if provider allows to timely

audits, since customer is the one who is ultimately going to be responsible of data
even when stored at SP [85]. The provider must allow customer to check if these
through third party audits and also use preventive measures need to be employed
for avoiding such violations [65][142][55][123]. Also the compliance can help the
providers to restrict the customer from violating regulations that are agreed upon
[109]. The author in [50] suggests to provide approaches for license management
which ensure a compliant deployment of cloud resources.
There are various issues discussed related, such as:
1. Data location [78] (as discussed in subsection 4.2.1).
2. Are service providers for cloud computing willing to support external au-
dits and security certifications, similar to those of the traditional service
providers [3].
3. Payment Card Industry Data Security Standard (PCI DSS), is a set of com-
prehensive requirements for enhancing payment account data security and
this was developed by PCI Security Standards Council to adopt consistent
data security measures on a global basis. This PCI DSS is a multifaceted
security standard, which includes requirements for management of security,
policies, procedures, network architecture, software design and other criti-
cal protective measures [3]. This is designed to protect customers account
data.
4. Traditional service providers are allowed to external audits and security

certifications. So, providers who refuse this are suggested to be used for
only trivial functions [178][22].
5. Virtualization and cloud computing have many issues that PCI Quality
Assessors (PCI QA), have concerns about and it is necessary to show com-
pliance to these rules, while deploying virtualization technology in PCI en-
vironment. Some of the important points to be noted are [3][28]:
a. Segregation of systems with one primary function per server and net-
work with isolation of all management and control networks
b. Virtual media that contains cardholder data needs to be protected.
c. Support auditing on system components.
d. PCI DSS may require additional processes or technology to ensure
patching and change control compliance in virtual environment.
e. Intrusion protection.
Guest operating system (OC1)

Generally an each customer in cloud is given a virtual private server (VPS),
which can run any operating system requested by user in a physical machine.
This means multiple operating systems can be running in a physical machine and
for a hacker find and trespass through any one of these operating systems running
on VPS is not hard. Its the user responsibility to keep constantly patching up
their own VM’s [105].
While interacting with CC services in some cases (where user is given complete
virtual machine such as in IaaS) virtual instances are completely under control
of customer. Amazon Web Services (AWS) does not give access to instances of
customers and so cannot log into the guest OS [146].
Host operating system (OC2)

Administration hosts (build specifically to protect the management plane cloud)
built to administer business needs in organization should be using multi-factor
authentication before giving access. When an employee quits from job or has
no longer access to the management plane, then his privileges should be revoked
[46]. This is because if a malicious user gets access to the host he/she can effect
the entire guest OS running in it which is a huge risk [105].
Instance isolation(OC3)
Instance isolation, Software isolation are different terms selected from primary
selection and discussed here.
When user access a CC service there is possibility for another user to be access-
ing same part of cloud or other part of it by creating its instance. Isolation is to
ensure that different instances running on same physical machine remain isolated
from each other. Since the administration of instances here is through instances
but not direct, it increases the risk and possibility of threat to the security of
CC users. Hence, there needs to be efficient system control and access control
restriction and a strict monitoring is needed to track changes. This isolation is
efficient in Xen hypervisor and is being used by Amazon [46].
Cost (OC4)
Cloud computing applications can be easily used, but securing these applications
requires management to spend additional resources (in terms of cash) [65]. Article
[44] discusses multiple challenges and inter-relates their success of implementation
to amount of cash an organization can spend.
Table 4.10: Other concerns that cater security of cloud computing

Auditability A
Compliance Com
Guest os OC1
Other concerns Host operating system OC2
Instance isolation OC3
Cost OC4
4.3 Identified solutions

These identified solutions are suggestions from different authors and could vary
from person to person or organization to organization. These solutions can be
used individually or in combination to give better solution for identified chal-
lenges.
Tables mentioned in section 4.4 (i.e., Tables 4.11-4.18), show the solutions
that are suitable for the challenges respectively (identified in section 4.2).
(S1) Encryption: Encryption is suggested as a solution to secure information

which is being transferred, stored (at rest) or under any other operation.
This section maps to the challenges identified in the previous section and
explains how these solutions can minimize the effect of challenges while
using cloud computing.
• Proposed solution for D1 (mentioned in 4.2.1) As data is at rest before

uploading into a cloud server to protect data from unauthorized par-
ties, data is suggested to be encrypted [144][146][132][155]. If data is
encrypted in IaaS, data-at-rest is possible and strongly suggested for
S3 [104], where as IaaS and SaaS cloud base application use Google
apps, or salesforce.com etc., which is feasible. Generally encryption
can’t be used for cloud based application because it prevents indexing
or data searching [104][78][44].
• If someone is sharing computing resources with other companies in a
public cloud, government may seizer data with reasonable cause and
might result in exposure of data. To avoid this and protect informa-
tion stored in public cloud data encryption is done, this avoids even
cloud providers from having access to data or decryption keys. If the
government or someone wishes to access data, a person needs to get
down to the user. This helps in maintain user’s data in private yet in
the same level for data access to cloud [46].
• To avoid access of data from other users Gatner states “applying en-
cryption on data that makes data totally unusable and normal encryp-
tion can complicate availability” [22][46][86].
• To avoid this problem (explained in section 4.2.2), it is necessary to
maintain multiple vendors and have Virtual Security Gateway em-
power. This is considered as the best proposed solution till date, which
will help organization get their own VPN network with full encryption
between all virtual resources around the world. By doing this they can
communicate and transfer the data by using encryption and security
[3].
• Data in SaaS vendors such as Amazon S3 is not encrypted by default
(to prevent problems explained in section 4.2.7, there is a requirement
to have the data encrypted before backup so that unauthorized parties
don’t get to access to such (backup) data [144].
• User-level encryption: Proposed as a solution to challenge specified in
4.2.8, in this problem to prevent data from being visible to unautho-
rized user, most commonly used solution is to have a user-level encryp-
tion, which is effective in cloud and for providing user-level encryption
there are tools and expertise already available [13][12].
• Encryption while transmission, the data should not only be transferred
to right person but should be transferred in from an authentic party
with total integrity maintained [92].
• Encrypting before uploading sensitive information into cloud storages.
Digitally signing and using RSA algorithm to encrypt is suggested [31].
• Cryptographic methods can prevent from losing data and protect data
integrity while transmission [100]. Symmetric and asymmetric cryp-
tography [75].
(S2) Access rights: Data owner should give permission to a particular party
so that they can access the data easily. To provide this data access control,
a standard based heterogeneous data centric security is used to give data
protection to application for preventing problems specified in sections 4.2.1
[146].
(S3) Use of central global transaction manager: Usage of central global

transaction manager refers to 2-phase commit protocol as per XA standards.
As a mix of on-premise and SaaS application can arise data integrity prob-
lems in the world of SOA and cloud computing (as mentioned in section
4.2.1. SaaS application’s functionality usually gets exposed to XML based
API’s. In SOA, many on-premise applications expose functionality based
on SOAP and REST web services protocols. There are standards avail-
able for data integrity in HTTP but are yet immature (WS-transaction and
WS-reliability) [144][123].
(S4) Ensuring boundaries: SaaS ensure that there must be clear boundary for
individual user data. This boundary must be ensured not only on physical
level, but also at application level to segregate the data from different users
(solution to data segregation and data access mentioned in section 4.2.1)
[144].
(S5) Following Electronic Communication Privacy Act (ECPA): Protec-
tion from government accessing electronic mails and other records done by
enforcing Electronic Communication privacy Act of 1986 (ECPA). When
this ECPA is applied in cloud computing activities, privacy or protection
is still unpredictable [144]. This is a suggestion for challenge on data con-
fidentiality issue 4.2.1.
(S6) Payment Card Industry-Data Security Standards (PCIDSS): SaaS
states it can provide better security (to customer’s data) when compared
to conventional methods. But even in such secure conditions the possibility
for an insider to still have an effect on customer’s data in a different way.
Insiders wont have direct access to database but still risk to security is
not reduced. To avoid this, SaaS vendor must be compliant with Payment
Card Industry-Data Security Standards (PCIDSS) where host merchants
must comply with PCIDSS. Which can be given as a solution to 4.2.1 [144].
(S7) Encryption using HMAC-SHA1: Before accessing data, first step is
authenticated using HMAC-SHA1 signature request with help of private
key [46][152][22]. By doing this the user can have full control over the data
to access which is a solution to problems in 4.2.1 [46].
(S8) Ship disks: For the problems mentioned in section (4.2.1, data transfer
bottlenecks), Jim Gray found the cheapest way is to send using low cost
ship disk. This will handle the case of large delay-tolerant point-to-point
transfer [13][12].
Example: if we want to transfer data of 10TB from U.C Berkely to Amazon
in Seattle, W.A. Garfinkel measured the bandwidth from S3 to three sites
where he found an average of 5 Mbits/sec to 18 Mbits/sec. for WAN link
20 Mbits/sec then
10 ∗ 1012Bytes 8 ∗ 1013
10T B = = seconds
20 ∗ 106bits/second 2 ∗ 107
= 4, 000, 000sec - approximately 45 days
Where ship disks can transfer 10TB data in one day that means like 1500
MBits/sec. AWS recently offering such a service called import/export
[13][12]. This could be used as a solution to data transfer bottlenecks.
(S9) Decommissioning and SSL encryption while transfer: AWS Proce-

dure decommissioning when data reaches useful life as stated in 4.2.1 AWS
uses a decommissioning process ensures that data is not exposed to unau-
thorized individuals.
Usage of SSL encrypted endpoints can provide more security to Amazon
S3. Using this method lets users access cloud features from Internet and
also from with in the Amazon EC2. There are certain test used to validate
level of security in SaaS, if there exist any vulnerabilities found out they
need to be given additional security 4.2.2 [144].
(S10) Usage of AWS and APL: To reduce possibility of denial of service (ex-
plained in 4.2.2) amazon AWS and APL are hosted large endpoints in inter-
net scale, world which can possibly reduce the chance of service downtime
[46].
For problem explained in 4.2.2, Amazon uses SSl protected end points to
authenticate AWS API’S and also generates a SSH host certificate, logs
them in instance’s console. This enables customers to use secure API’s
before logging into instance for first time and to make transactions more
secured customers are forced to use SSL for interaction with AWS [46].
(S11) Acceptable Use Policy (AUP): Solves the problem of port scanning
mentioned in 4.2.2, it is required to maintain a policy; every action of
customer need to be monitored and any violations by customer should be
reported to the service provider. Amazon EC2 maintains a ‘Amazon EC2
Acceptable Use Policy (AUP)’ which checks for any policy violators and
warns service provider in early stages [46].
(S12) Re-evaluating models and using enhanced security mechanisms:

To get viable solution for set of issues mentioned in section 4.2.4 and for
utilizing potential benefits of cloud, existing life cycle models, management
processes, service attestation, penetration testing and risk analysis must
be re-evaluated. Security mechanisms must be upgraded to use Security
Assertion Markup Language (SAML), Extensible Access Control Markup
Language (XACML), and Web services standards [6]. For policy engineer-
ing, there are several approaches proposed in cloud adoption, which may
cause disruptions to the organization [175][146]. To avoid this disruption set
of roles should be proposed for existing and optimal sets. Another possible
solution is StateMiner approach that helps to get heuristic based solutions
on RBAC state with least structural complexity [146][145][6].
(S13) Usage of standards: This is a solution/suggestion given to identity and

access management system (described in section 4.2.4) and using this system
(cloud computing system, maintaining identity and access management)
cloud service providers are suggested to adopt standards such as SPML,

SAML [78], open id standard [78], OAuth, and XACML to have a secure
interaction between entities (users of a system) in different cloud computing
applications [6]. Entities in cloud platform could also be identified and
granted access rights based upon delivery models (SaaS, PaaS, IaaS, etc.),
and type of cloud (hybrid, public, private etc.), to ensure better security
[123]. A comparison between different ways of identity management and
standards for identity management are discussed in [66]. The authors for
[85] say SAML is better than OAuth.
(S14) Disaster recovery and Business continuity: For safeguarding data

from unplanned emergencies or minimizing enterprise downtime (the effects
of these are mentioned in 4.2.4), appropriate plan for Business Continuity
(BC) and Disaster Recovery (DR) is suggested [144]. For the problem men-
tioned under failure in providing security (4.2.4), is to have a system which
restores data from backups silently, without any interruption and this sys-
tem called failover [127].
(S15) Maintaining multiple stacks: is a solution for maintaining availability

(challenge description in 4.2.4), is to maintain multiple service providers
providing different stacks, such that if one stack fails there will be another
stack (i.e., service provider) supporting as a substitute [13][12].
(S16) Trusted Third Party: Another solution to maintain availability (chal-

lenge discussed in 4.2.4), is by securing the system by using Trusted Third
Party (TTP) [179], when a trusted third party is used it provides the
strength that all organizations involved in its activities are trustworthy [94].
To ensure policy management and trust management (for problems men-
tioned in section 4.2.4) is to have a Trusted Third Party (TTP) within the
cloud to ensure confidentiality integrity and communication [179].
(S17) Keeping track of users: Managing multiple SaaS products may also
increase IT management overhead and might also complicate the user man-
agement system and in section 4.2.4 explains a scope of misusing user infor-
mation if not managed properly and to avoid this issue, customers should
keep a track of persons leaving or joining the organization, correspondingly
add or remove their information [144].
(S18) Privacy preserving protocols: Solution to the problem of authentica-

tion and identity management (mentioned in section 4.2.4), is to have pri-
vacy preserving protocols, which verifies various attributes. An example of
such technique is zero-knowledge, proof-based technique, these techniques
use pseudonyms and accommodate multiple identities to protect privacy of
users and can further help built a user centric federated IDM for clouds
[146].
(S19) Dynamical scaling: As a suggestion to the problem mentioned in section

4.2.5, open cloud manifesto states that stay away from problems with allo-
cation of resources during peak hours cloud should be dynamically scalable
(scale up and down) [61].
(S20) Virtual Security Gateway (VSG): Virtual Security Gateway (VSG)

which is proposed as a solution to the challenge mentioned in section 4.2.3,
is a firewall that runs inside virtual infrastructure and ensures all communi-
cations inside virtual environment stick to defined security policies. In this
infrastructure no two virtual machines are allowed to talk to each other
without passing through the security gateway. This gateway uses VPN
encryption to ensure secure interconnection between VM’s, since VSG is
placed inside virtual infrastructure auditing can also be achieved, regula-
tory compliance requirements can also be satisfied. As the cloud computing
platform is scalable user can extend his/her services easily, which means
this VSG can be easily deployed. VSG also allows creating lab/test envi-
ronments and thereby improving overall security [3].
(S21) Alter this to software: This is given as a solution to a problem men-

tioned in 4.2.3, this can be done by maintaining a hypervisor level, MAC
and trusted computing techniques are suggested mechanisms used to build
future secure cloud systems [73].
(S22) Including auditability as a layer: As a solution to the problem ex-

plained in auditability (section 4.2.8), this could be added as an additional
layer that stays out of virtualized guest operating system’s reach, which
provides arguably more security than those built into the application. Such
new features reinforce the cloud computing perspective of changing focus
from specific hardware to virtualized capabilities [13][12].
There are some well known auditing guidelines such as in SAS 70, SOX and
HIPAA. US government follows FISMA, NIST and FIPS [35].
(S23) Multi factor authentication: The solution recommends for challenge

in this section 4.2.8, is to remove traditional password based encryption
and use some multi-factor (two ways of checking before providing access)
to provide access at least use a certificate-based SSH Version 2 authentica-
tion to access their instances [147][145][21][146]. The author explains with
an example such as if someone using Linux as his/her guest OS, hardens
the instance they should use certificate based SSHv2 to access the virtual
instance, disable root login, use command line logging and ‘sudo’ command
to change privilege [146]. General methods of access control are use of roles
and password, but using a multi-factor authentication is more secure than

single password protection [105].
(S24) Permission revoke: Problem mentioned in 4.2.8 gets solved if permis-

sions are revoked as soon as an employee quits from his job or no longer
has access to management (which he had before) [146].
(S25) Xen hypervisor: This solution is suggested for a challenge mentioned in

4.2.8, which mentions that to effectively isolate instances there is needs for
an efficient system control, access control restriction and a strict monitoring
that tracks changes within virtual machines. It is also stated that isolation
is efficient in Xen hypervisor, which is currently being used by Amazon
[146].
(S26) Calculate hash: This is a solution proposed for backup related issues
(section 4.2.7). Before uploading a file, the client is suggested to verify
whether the file is stored on backup drives or not during the intermediate
transaction and whether keywords in file remain unchanged. Calculate the
hash of the file before uploading, which also divides file into equal size
randomly and stores it in the locality [42].
A general solution in providing security to the cloud computing system is by
having a service which acts as a watchman and secures all the applications
designed in it [88].
For storing data on the cloud there might be some issues, as mentioned in
4.2.1. Storing hash in trusted local memory is a method to avoid issues,
it helps user to check if the data stored in cloud stays unaltered. If there
are huge sets of data is being stored, individual can maintain his own hash
table/hash tree. This mechanism is used in many storage prototypes such
as TDB [26][102], SiRiUS [26][59] and is commercially being used in Solaris
ZFS filesystem1 [26].
This can also be used for data integrity but maintaing hash can become
complicated [81].
(S27) Credential or attribute based policies: Using credential or attribute-

based policies is a suggestion to identify unauthorized users [146].
(S28) Standardize API’s: Suggestion to the problem with data lock-in (4.2.1)
is to standardize API’s, such that customers can deploy services and data
across multiple cloud computing providers. The advantage of doing this
even if one service provider fails it would not corrupt all other copies stored
on other places (data or services). If this format is introduced, it could
create a new usage model where software structures can be used in an
internal data center and in public cloud [13][12].
(S29) Isolate personal information and software: For problems mentioned

in privacy (i.e., section 4.2.6) author in [29] proposes that to maintain pri-
vacy of an individual’s data a solution could be to isolate users private
data in databases and distribute the software around the internet. Author
in [65] suggests using software isolation where each domain in the cloud
environment has a trust agent.
(S30) Middleware for configuring storage infrastructure: Cloud comput-

ing providers often use Storage Area Networks (SAN) to gain storage scal-
ability, which have a vast parameter range and hence makes manage cloud
storage, components to configure. To solve this problem a solution was
proposed which introduces a SAN configuration middleware that manages
task of updating and debugging heterogeneous SAN developments [47].
(S31) SaaS protection: Homomorphic token with distributed verification of

erasure-coded data towards ensuring data storage security which supports
dynamic update, delete and append data without loss or corruption. This
method is efficient over byzantine failures and server colluding attacks, data
modification [65].
(S32) Border Gateway Protocol (BGP): To identify malicious systems that

announce themselves as the righteous destination for all the data transferred
in a network. This architecture is vulnerable to DDoS attacks [65].
(S33) Quantitative risk assessment framework (QUIRC): Author in [136]

mentions that this framework can be used to assess the level of security in
the enterprise. The usage of this framework is suggested by [17]. In addition
to this wide-band delphi model is also suggested [17].
(S34) Distributed access control: Author [8] suggests distributed access con-
trol architecture for access management in CC.
(S35) Intrusion severity analysis: Author [14] identified intrusion severity

analysis in CC is a problem and for that he proposed a new method to
detect intrusion, which is based on machine learning.
(S36) Performance oriented architecture: This is a reactive but not proac-

tive architecture, which analyses the CC intruders before they can cause
any damage [113].
(S37) Trusted virtual data center: These data centers have different VMs
and associated hardware resources which will help to identify, which VM
will access which resource. This can also separate customer workload to
different associated virtual machines [74].
(S38) Attack detection model: Article [90] proposes a proactive attack detec-
tion model which can detect an attack (at time of attack or before), alert
the system/security administrator and helps the customer to understand
the kind of attack that happened by looking at pattern of attack. Solu-
tion to attacks from fraud pissing or denial of service etc., is to keep login/
account credentials personal and never disclose them.
(S39) Access control based on group signatures: By this method user can
access CC using a group account, instead of personal account. This will
allow the customer to preserve his/her privacy [176].
(S40) Proactive approach for CC security: A proactive approach to prevent

customer falling into wrong hands is being proposed. This model proposes a
security cloud which can perform tasks such as: Monitoring public blacklists
for ones own network blocks, Vulnerability Assessment, Penetration Testing,
Log Analysis, Host Based Intrusion Prevention System [142].
(S41) Efficient remote data possession checking (RDPC): The author for
article [33] mentions that this scheme is better in terms of communication
and computation, verification without need to be compared with the original
data and mentions that user needs to store only two secret keys and several
random numbers.
(S42) Non-repudiation protocol: A nonrepudiation protocol is proposed. This

protocol is a two-party non-repudiation protocol that is like TCP/IP 3 phase
handshaking protocol. This protocol used for both integrity checks and non-
repudiation [51].
(S43) Two stage API access control: The author of [139] proposes a two
stage API access controlled mechanism that developed based on a Rule
Based Access Control mechanism (RBAC).
(S44) CISCO secure data framework: This framework provides multiple se-
curity layers with different existing security technologies that can be used
to enhance security [169].
(S45) Security measuring framework: This framework is suitable for SaaS

and can be used to determine the status of user’s applications that have
been running in the VM for some time [75].
(S46) Community watch service: This service is supposed monitor, analyze

cloud users and detect newly injected malware attacks [75].
(S47) Network based intrusion prevention system (NIPS): The author of

[75] gives this a multi-technology based approach. This NIPS has 4 different
technologies clubbed together such as, hardware acceleration, active defense
technology and linkage with a firewall and a synthesis detection method that
are used in combination to detect block visits when threats are detected in
real-time.
(S48) Data fragmentation: This solution is suggested to prevent intrusion into

user’s privacy. The suggestions are to use a cloud based malware scanners
and use data fragmentation technology which can be used to hide the con-
nection between data attributes and/or separate customer’s data from the
software [75].
(S49) Virtual network framework: Using a virtual network framework which

consists of routing, firewall, shared network and ability to control inter-
communication of VMs deployed, can provide secure access at VMs level
[75].
(S50) SaaS Role Based Access Control (S-RBAC) and Intercloud iden-
tity management infrastructure: Are two methods that the author of
[75] mention which can avoid the conflicts with access control while working
with a multi-tenant environments.
(S51) Data protection as a service: This service model is proposed to ensure

privacy and data security, also offers evidence of privacy to owners [140].
(S52) Confidentiality as a service (CaaS): Proposes to protect the visibility

of even clear text data from unauthorized users by having multiple cumu-
lative layers of encryption [48].
(S53) Proposes a data security model: This model consists of 3 layers which
perform their own task. First layer OTP authentication, second (data en-
cryption, data integrity, private user protection), third fast data recovery.
Test this model and implement software to improve work of data security
model in CC [15][164].
(S54) Third party auditing scheme: This is to protect the data storage secu-
rity and proposal here is to move the auditing process into cloud computing
and integrate with the architecture of cloud provider [68][32] [81]. Public
and private auditing [153].
(S55) Remote data position with public verifiability: Proposes Remote

data possession checking protocol with public verifiability. This uses HLAs
and RSA construction [98].
(S56) Fine grained access control: Proposes a fine grained access control
mechanism for data [95].
(S57) Fog computing: This method uses behavior profiling of user and a decoy
which can prevent malicious users from damaging the CC environment [143].
(S58) RSA based Storage security (RSASS): This method is supposed to

address remote data security, which is based on RSA for storing files in
remote servers. This method can also compute large files with different
sizes [153].
(S59) Chinese wall policy: This method is used to address the insecure infor-
mation flow at by using information flow policies at IaaS layer [163].
(S60) Privacy aware inter-cloud: This method takes charge of data protection
standards, processing the privacy requirements and user requirements [149].
(S61) Multi cloud model: An OPNET tool is suggested here which has sepa-
rate security as a service and tested simulation of all traffic [5].
(S62) sTiles:This is to preserve the privacy from both inside computational

nodes and intruders [23].
(S63) Trust evaluation model: This ETCT architecture proposed includes

time variant and space variant trust evaluation methods for calculating
direct trust and recommends trust [64].
(S64) Multi-replica PDP model: This model is proposed to preserve integrity

and data availability in the cloud. The authors of [99] proposed an improve-
ment to this model, which can support third party auditing also.
(S65) Mutual protection for cloud computing architecture (MPCC): Is

designed on a reverse access control concept which gives user the control for
authorization and authentication; also allows to check if any CC security
violation are present [7].
(S66) Network protection methods: Using SSL/TLS, IPsec, Traffic cleaning

and network based intrusion detection [32].
(S67) Honeypot: Used to detect and counter attack or deflect unauthorized

access attempts [32]. This can be loaded into the user’s logical instance
running on SaaS, PaaS or IaaS and help them to effectively handle viral
infections and service hijacking in the cloud.
(S68) Preserving cloud computing privacy: In [122] authors of this paper

propose a PccP (preserving cloud computing privacy), which prevents unau-
thorized users from accessing user information or derive such information
from IP addresses.
(S69) Permission as a service (PaaS): This solution is proposed in par with

other solutions, which allow users to tell what part of data can be accessed
in which way, but the problem here is when there are large chunks of data
to be managed. So to provide a solution the authors designed a new service
which separates access control panel and allows user to set access to all the
data from a single location. This PaaS uses Attribute Based Encryption
(ABE) in their framework [45].
(S70) Tree based key management: Existing tree based encryption methods
have some weaknesses mentioned in [177]. So to avoid this weakness and
propose a practical solution for private data management, the authors de-
velop a new tree based key management system and prove that the above
problem can be solved [177].
(S71) Byzantine fault-tolerant protocol: This byzantine fault-tolerant repli-

cation protocol is used to manage collaborative data storage. Zetta which
is similar to RAID systems can also be used as a solution [100].
(S72) Denial of service prevention methods: Network access control (NAC),

Intrusion Detection System (IDS) and Intrusion Prevention System (IPS)
[100]. Intrusion detection system (IDS) or firewall [67].
(S73) Computational Intelligence (CI): CI is a mathematical model for CC,

and authors of [57] propose an improvement that can enhance the perfor-
mance and automated service provisioning in data security.
(S74) Ubiquity and integration of services (UBIS): The new architecture,

UBIS, discussed in article [67] is proposed to satisfy cloud user requirements
and cloud security challenges. This architecture is based on QoS manage-
ment and includes security aspect based on an event driven architecture
and ubiquitous services in the cloud.
(S75) Risk perception model: Author in [49] proposes a model useful in defin-
ing risk perception before moving on to CC. This model was redesigned to
reduce the 9 variables (such as, voluntaries, immediacy of effect, knowledge
about risk, knowledge of science, etc. that were used by Baruch Fischoff
[53] to determine when a product is safe to be accepted) into two areas
(understanding and consequences). The perceived risk in this model is a
function of these two terms understanding and consequences.
(S76) Proxy based firewall/NAT traversal solution (PASS): This solution

is proposed for SaaS integration, which allows on-premise applications to
work securely on CC [97]. Other solutions are VPN, which is difficult in
SaaS but easy on IaaS, Microsofts AppFabric (specific to Microsoft plat-
forms) and TARGET.
(S77) RSA based Assumption data integrity check: This method is pro-
posed by [81] since other methods such as hash, 3rd party auditing etc.,
he found are complicated or insecure. This new method combines both
identity based cryptography and RSA signature [81].
(S78) Capability based approach: Use of Access Control List (ACL’s) or file
groups lack scalability, fine grain access control in cloud computing. The
scalable method (proposed in article [170]) that is based on using the com-
bination of ABE, lazy re-encryption and proxy re-encryption is supposed
to become complicated in an extremely dynamic environment such as CC.
The re-encryption scheme and framework for access control on published
XML documents by using different cryptographic keys on XML documents
are also not suitable. Since XML scheme becomes complex to manage keys
and XML documents and in case of re-encryption, if one user is malicious
the data is exposed. So the author for [133] proposes a scalable, efficient
and secure data access control mechanism using Capability based access
control.
(S79) Improved Capability based approach: The author suggests to use a

capability based approach which is already used in distributed file systems
and suggest some extensions to them. The extensions constitute integra-
tion with external identity or access management components used by en-
terprises, capability based access control architecture for data centers, user
to user and user to application delegation and fine-grained access, dynamic
access right’s scope (a survey on the existing methods by the authors show
that the features they proclaim to have in the proposed architecture are
missing within the existing ones) [70].
(S80) Multi tenancy for databases: Proposes a multi tenant database design
for ad-hoc clouds. The authors boast that their method allows multiple
organizations to collaborate and benefit without security leaks [118].
(S81) Secure ranked keyword search: Searching through encrypted data in

CC using traditional systems could accidentally reveal data, this proposed
system avoids leaking others data [156]. This proposed method is experi-
mentally tested.
(S82) Security framework: The author in [148] proposes a comprehensive se-

curity framework for CC, which has separate modules for trust issues of key
CC components and also a module to handle security.
(S83) Automatic virtual infrastructure evaluation: The author proposes

an approach to automatically check if the virtual infrastructure is following
the security requirements imposed by the user [56].
(S84) Multi-tenant authorization model: This model is suitable for middle-

ware services in PaaS layer. This model is said to support RBCA, Multi-
tenancy, path based object hierarchies, hRBCA, and federation. Some other
methods of authorization are also described in related work. But this model
is based on abstraction of path concept and the authors mentions that it is
better suited to CC authorization [27].
(S85) Data-driven framework: The author for [178] proposes a data-driven

framework which provides secure data processing and sharing between cloud
users. This framework is designed in such a way that secure query processing
exists in multi-user scenario, integration of declarative access control with
data processing, efficient end to end data verification and system analysis
and forensics by capturing accurate historical records exists. This model is
ready to use and is also available for download.
(S86) Clout Trace Back (CTB): This model is being proposed to trace back
the DDoS attacks in CC [83]. It helps to find the real attacker among the
number of requests that server receives under DDoS attack state [83].
(S87) Software watermarking: This paper designs a robust and new cloud
based watermarking mechanism that can report and alert the required per-
sonal [171].
(S88) Tag: Tag entities which should not be moved and prevent CSP from mov-
ing sensitive data to another jurisdiction [71].
(S89) Fine-grained access with delegation of computational intensive

tasks: This model achieves fine grainedness, scalability and data confiden-
tiality for data access control in cloud computing. This proposed scheme
enables the owner to delegate most of computation intensive tasks to cloud
servers without disclosing data contents or user access privilege information.
The proposed scheme is provably secure under the standard security model
[170].
Apart from the above mentioned solutions some of the authors have mentioned
general guidelines to provide security:
• Article [15], Good and well formulated SLAs are to some extent helpful
points for ensuring availability. To avoid risk of insider malicious attacks
the user must be informed about the information security and management
practices. To prevent outsider attacks there can be a network access control
system built up to at OS level. VMs should be isolated for each customer.
Stopping to share account credentials and two factor authentication proac-
tive monitoring (detect unauthorized activity) will probably help service
disruptions. A strict registration and validation can reduce a lot of damage
from attacks. Defense-in-depth approach at different layers using different

protection mechanisms can protect the VM’s better. Being aware about
threats and preventing sensitive data into cc can be helpful in case of user
and providers should allow the users to port their security protocols so that
the user’s data stays protected.
• Article [105], Settlement on open standards can help reduce vendor lock-in
problems and incompatibility issues. While transferring data user needs to
use a SSH-tunneling or VPN protocol. The user can look for CSP with
continuous data protection where every version of file is saved by the CSP
in their server. Data portability should be provided to avoid Vendor lock-in.
To avoid DoS the simplest way is to check if the source ip-address is invalid
and another is to implement a strong and robust firewall. A reverse firewall
can help to protect attacker from accessing an external server from a VPS.
• Article [89], The author highlight some issues that can build up a trust in
cloud computing platform:
1) Notify the customer when any entity accesses data,

2) Without informing the customer the SP won’t save any backups and
3) When a customer withdraws from Cloud all his data, direct and indi-
rect access to that data will be destroyed.
In addition to this the software provided should be reliable, transparency

at service level security properties and where the data is stored (its phys-
ical location). Some emerging technologies such as remote access control,
reflective property, certification and private enclaves should be maintained
so that the trust enhances.
• Using CSA suggested security solutions [137].
• Article [54], Providing tools for security management and auditing, disclos-
ing security policies, compliance and practices can attract people towards
cloud computing. Cloud providers need to have strong internal monitoring
and auditing scheme.
4.4 Results of literature review

Literature review on Cloud Computing (CC) research area resulted in numer-
ous challenges related to security (43 challenges), which were categorized and
explained. Following tables summarize the findings of literature review. The ta-
bles 4.11-4.18 show grouping of challenges, their respective code and number of
solutions and the number of research papers that discuss the challenges.
As an example, from table 4.11 which shows ‘data related’ challenges their
codes, solutions identified and number of references; we can identify that the
challenge ‘data security’ which is given the code ‘D1’ (mentioned in ‘data related’
challenges) has ‘(S1), (S2), (S79), (S73), (S53), (S51), (S81)’ as solutions with
‘27’ articles mentioning this challenge in their discussion.
Similarly other tables ‘4.12’, ‘4.13’, ‘4.14’, ‘4.15’, ‘4.16’, ‘4.17’ and ‘4.18’ shown
below mentioned about network, virtualization, organization, scalability, confi-
dentiality and privacy, backup and other concerns related challenges respectively.
The summary includes corresponding codes (used to mention the challenge), so-
lutions (identified from literature which is a mitigation strategy for the identified
challenges) and number of references (number of articles mentioning the chal-
lenge).
Table 4.11: Data related challenges and solutions
Relation Description Code Solutions No of

refer-
ences
Data Security D1 (S1), (S2), 27
(S79), (S73),
(S53), (S51),
(S81)
Data locality D2 (S1), (S88) 14
Data related Data integrity D3 (S3), (S54), 19
(S1), (S87),
(S26), (S77)
Data segregation D4 (S1), (S4), 6
(S34)
Data access D5 (S4), (S27), 12
(S89), (S1),
(S78), (S56),
(S69) (S79)
Data confidentiality D6 (S5), (S85) 4
issue
Data breaches D7 (S6), (S44) 5
Reliability of data D8 (S41), (S55) 7
storage
Data center operations D9 (S7), (S8), 14
(S71), (S1),
(S53)
Data sanitization D10 (S7), (S8) 9
Data storage D11 (S28), (S26), 35
(S54), (S42),
(S31), (S58),
(S80)
Table 4.12: Network related challenges and solution

refer-
ences
Network security N1 (S1), (S9), 20
(S35), (S36),
(S66), (S40),
(34), (S47),
Network related (S46)
Application N2 (S13), (S43) 15
vulnerabilities
Host and network N3 (S36), (S72) 4
intrusion
Denial of service N4 (S10), (S86), 13
(S72), (S38)
Men in middle of attack N5 (S10) 6
Ip spoofing N6 (S32) 3
Port scanning N7 (S11) 2
Packet sniffing N8 No solution 2
Sharing computing N9 No solution 11
resources
Table 4.13: Virtulaization related challenges and solutions

refer-
ences
Vulnerability in V1 (S37), (S83) 17
virtualization
Virtualization Related Virtual machine V2 (S20) 5
protection
Networking in virtual V3 (S59), (S49) 10
networks
VM isolation V4 (S25), (S29) 9
VM securitization V5 (S21) 1
Table 4.14: Organization related challenges and solutions

refer-
ences
Organizational security O1 (S12), (S75), 8
management (S12), (S76),
Organization related (S33), (S45),
(S74), (S82)
Trust O2 (S16), (S63) 17
Failure in providing O3 No solution 6
security
Identity and access O4 (S13), (S17), 43
management (S18), (S34),
(S79), (S57),
(S84), (S59),
(S50), (S61),
(S65), (S67),
(S84)
Availability O5 (S14), (S15), 21
(S16)
Table 4.15: Scalability related challenges and solutions

refer-
ences
Incompatibility issue SC1 No solution 19
Scalability related Constant feature SC2 No solution 1
addition
Scalability SC3 (S19) 4
Table 4.16: Confidentiality and privacy related challenges and solutions

refer-
ences
Confidentiality C1 (S52) 9
Confidentiality & privacy
Privacy C2 (S29), (S62), 28
(S48), (S39),
(S60), (S68),
(S70), (S64),
(S81)
Table 4.17: Backup related challenges and solutions

refer-
ences
Backup B1 (S1), (S26) 10
Retention or backup
Data retention or B2 No solution 13
Recovery
Table 4.18: Other concerns related challenges and solutions

refer-
ences
Auditability A (S1), (S22) 8
Compliance Com No solution 14
Guest os OC1 (S1), (S23) 2
Other concerns Host operating system OC2 (S24) 2
Instance isolation OC3 (S25) 1
Cost OC4 No solution 2
From Tables 4.11-4.18, it can be be understood that in most cases identified

challenges have at-least a single paper describing issue, but every challenges iden-
tified does not have a solution or practice proposed. From tables, its clear that
there are no solutions found for some challenges. Given below are list of those
challenges:
1. In Network related challenge N8 (‘Packet sniffing’), N9 (‘Sharing computing

resources’).
2. In Organization related challenge O3 (‘Failure in providing security’).
3. In Scalability related challenge SC1 (‘Incompatibility issue’), SC2 (‘Con-

stant feature addition’).
4. In Backup and recovery B2 (‘Recovery’).
5. In Other concerns COM (‘Compliance’), OC4 (‘Cost’).
Out of these challenges with no solutions, ‘Incompatibility’ under scalability

challenges with ‘SC1’ as code is a challenge that has 19 articles mentioning the
challenge and does not have any proper mitigation strategy mentioned.
Chapter 5
Surveys: Design and results
The results that are obtained from chapter 4 conclude that there are some chal-
lenges with no mitigation strategies proposed. These are ‘packet sniffing’, ‘sharing
computing resources’, ‘failure in providing security’, ‘incompatibility’, ‘constant
feature addition’, ‘data retention or recovery’, ‘compliance’ and ‘cost’. Out of
these (challenges with no mitigation strategies) incompatibility is a challenge
mentioned by a number of articles but no mitigation strategies identified. To
identify mitigation strategies for the challenges, the surveys are employed.
This chapter describes the steps for designing the surveys, identify the survey
respondents and implement the surveys.
5.1 Survey design

Surveys are described as the best method to collect the original data in a pop-
ulation that is too large to observe directly [158]. Surveys generally help the
researcher to understand and generalize the results on a population by asking
questions to a sample from that sample [36]. Surveys are chosen over other meth-
ods because these can be simple and in this case since the objective is just to
identify solutions/practices/guidelines for the challenge that has no mitigation
strategies, surveys stand a better chance in reaching large population. Three
types of surveys are reported [158] and as follow:
• Needs assessment survey: to ask for problems and gather possible solutions.
• Marketing survey: surveys to understand the demand and the nature of

specific programs.
• Evaluation surveys: to evaluate the outcome of newly implemented pro-

grams and policies.
These surveys can be conducted in different ways, by:
• Conducting surveys through mails i.e., by sending a set of questionnaires to

the respondents with a covering letter which describes the purpose of the
survey.
86
Chapter 5. Surveys: Design and results 87
• Dialing a random number from a sample selected from a telephone directory

and registering their responses to the questions either by taking a note on
paper or by entering them into a survey form.
• Face-to-Face surveys are another way of gathering responses by asking the

respondent a question. These kinds of surveys are expensive and time tak-
ing.
• Drop-off survey where the questionnaire is given in hand to the respondent.
• Electronic surveys are the kind of surveys where electronic media are used
to contact the respondents. The respondents are sent a set of questions
through email and requested to respond.
The needs assessment type is considered in this work to gather possible solu-
tions/practices/guidelines. The medium used to communicate with the respon-
dents is using e-mail and the responses are collected using an online survey tool.
5.1.1 Steps for conducting an online survey

For conducting an online survey the steps similar to other methods such as plan-
ning, data collection, data analysis, reporting and application need to be followed.
Each of this general procedure has few specific tasks to be followed [129].
• Identify the evaluator: The evaluators are the master students of their
master thesis report under the guidance of the supervisor.
• Identify and engage the stakeholder: The stakeholders are those who have
an interest in the program (results of the thesis). In this context the stake-
holders are the company professionals who could use the benefit of working
with multiple CSP’s provided issue with incompatibility will get solved.
These can be solutions architects, senior practice managers, chief archi-
tects, people who have rigorously worked with cloud computing application
deployment, founders and CEO’s. The survey process used in this study
identifies these stakeholders from different companies and ask them to re-
spond to the survey questions.
• Determine resources: The resources is based on the respondent’s time spent

on the survey link, which is generated from a free online survey tool. The
results can also be analyzed with the help of an online survey tool. This
online survey tool generates a report, which summarizes the responses for
each question presented in the survey questionnaire.
• Writing goals and objectives: The survey aims to collect expert opinion and
experiences on how to mitigate the issues identified under ‘incompatibility’.
Since there were no dependable mitigation strategies (which can be used as
a mitigation strategy by companies implementing CC) identified from the

literature review. The incompatibility issue identified in the preliminary
study was broken down into 3 issues for easy interpretation in survey ques-
tionnaires. These 3 issues can be stated as objectives of performing surveys,
which were clearly stated in the beginning of the survey.
• Evaluation or results: The evaluation of results will be done in such a

way that, the results mentioned through this study can help new CC users
to identify mitigation strategies for the incompatibility issue. For the re-
searchers this evaluation will help to identify which areas must be concen-
trated to improve security for CC users.
• Using a software to implement surveys: The survey link is created at (www.

quicksurveys.com). The survey link is kept active to gather as many
responses as possible that can suggest mitigation strategies. The survey
link is also a one time response link which assures that there won’t be
multiple responses from the same person (this avoids duplication) and this
can improve the quality of surveys.
5.1.2 Designing the survey questionnaire

To see that the results of the study are helpful for the research field, there are 5
questions which were mentioned in [158] and have been taken into consideration
while designing questions. The questions are as follows:
• What do we want to know?
• About whom do we want to know?
• How do we word the questions?
• How do we elicit appropriate and adequate responses?
• How do we interpret the results?

The above mentioned 5 questions were kept into consideration and survey
questionnaire were designed. These were designed to focus on the challenge
that has no mitigation strategies identified from the literature. The questions
in the surveys were framed taking suggestions from our supervisor and we con-
stantly simplified (checked and corrected) to avoid grammatical errors and help
better understanding for the respondents. The final questionnaire used for sur-
veys are presented in the appendix B. The survey questionnaire created fol-
lowing the above mentioned procedure was open for answers through the link
(http://tolu.na/17cFxZz). The respondents were contacted through e-mail
and requested to fill out the surveys. The respondents were also sent reminders
at regular intervals to make sure that the responses were received without delay.
It was also necessary that the survey responses are strong (number of responses
gathered from experienced practitioners) because the sample size considered was
random. The survey link was kept open till 15 filled responses (with responses
from experienced personnel) were received.
5.2 Observations from survey responses

Given below is the description of responses observed from surveys. There were
27 survey visitors and 15 responses. Among these, there were responses from
CEOs, Founders, Chief architect, solution architect, senior technology practice
manager, etc. These respondents are Information Technology (IT) professionals
having 8, 7, 5, 3 years etc., experience in cloud computing and from different
companies. It is also possible that some of these respondents were working with
cloud computing from the year it was being used. Given below is the discussion
of responses received for each question.
5.2.1 Results: Question related to service providers, ser-

vice models and cloud types used
There were few basic questions asked to the survey respondent such as, who is
your service provider, which cloud service model and deployment model they use.
These questions were asked to know which service provider, which service model
and deployment model are popular in practice.
Q1: Who is your service provider for cloud computing applications?
The answers to this question are shown graphically in figure 5.1. From this
figure its understood that Amazon is the leading service provider with 60% votes,
followed by Google with 33.33%.
Figure 5.1: Results of survey: Service providers used

Q2: Types of services used

In figure 5.2, it can be understood that SaaS is being predominantly used
(with 66.67% votes from respondents), followed by IaaS (with 46.67% votes from
respondents) and in the last comes PaaS (with 40% votes from respondents).
Figure 5.2: Results of survey: Type of service used
Q3: Types of cloud used

In figure 5.3 shows the types of cloud are being used. From this it can be
understood that public clouds are mainly being used (with 66.67% votes from
respondents), followed by private cloud (with 46.67% votes from respondents),
the next position is hybrid cloud (with 26.67% votes from respondents) and in
the last comes community cloud (with 13.33% votes from respondents).
Figure 5.3: Results of survey: Deployment models used
5.2.2 Results: Questions in relation to cloud computing

incompatibility issue
For the question related to ‘incompatibility’ in cloud computing systems that are
presented in this questionnaire, the responses were guidelines and precautions.
Some of these suggestions were to use global standards to provide interoperabil-

ity between service providers and different ways of encryption while interfacing
between multiple cloud providers.
It is also observed that there have been maximum responses for the first 3-4
questions and for the last question related to on premise authentication systems
there were few responses. For easy understanding each respondent of the survey
will be given a notation such as R1, R2, R3 so on till R15 to describe respondent
1, respondent 2, respondent 3 so on till respondent 15 respectively. Given below
are the responses of respondents to each question.
First question:
a) What steps can be followed to provide interoperability between

two or more service providers?
Responses: Total Respondents to this question are 13. Given below are the
solution/practice/guideline mentioned against each respondent.
R14 (1) Gateway functions.

(2) Identity aware applications
(3) Federated authentication and Pass through mechanisms.
R8 (1) Check compatibility of software
(2) Check upgrade path of both the softwares
(3) Have schemas and architecture very clear as how do you want to
use data flow between them and how do you want to see it as a
single solution rather than 2 softwares.
R12 (1) Avoiding synchronous communication between clouds as much as
possible by engaging an acquire-store-resend model.
(2) Monitoring connections in the integration hub at all available lev-
els by reserving a mechanism for an automated acquiring of lost
connections.
(3) Putting the maximum attention on semantics and ontologies of
operations and data involved in the interactions between clouds.
(4) Minimizing the number of interactions between clouds
R4 Cloud computing vendors have formed a common platform cloud com-
puting interoperability forum (CCIF) to address the problem of cloud
interoperability and standardization. Unified cloud computing is try-
ing to unify various cloud APIs and abstract it behind an open and
standardized cloud interface. Thus a key driver of the unified cloud
interface (UCI) is to create an API about other APIs
R2 Common Standards
R5 DK
R6 Flexible licensing options

R7 Follow the proper regulations, security, standardization
R9 It depends upon the specific scenario and would be driven by the appli-
cations which have to be interoperable. As of now the IaaS and PaaS
providers are still evolving rapidly so thoughts around integration are
premature. Bigger challenge is for a market facing service provider
who has to offer a bouquet of services to the end customer and these
services are hosted on different platforms.
R11 Standard interfaces and security Inter-operable software Cross-platform
communication software This would be best achieved by following a
global standard in implementing cloud platforms. But I doubt if such
a thing exists.
R3 Two or more systems or components are used for exchanging informa-
tion to provide effective and efficient exchange between more service
provider components like SAS, SSIS.
R13 We deploy a platform/interface for management/orchestration be-
tween different service providers.
Observations:
From the answers that are given by the responder to question one, it can
be understood that the responses to this question are mostly guidelines. 3
out of 13 respondents for this question say that using standardization will
be better option and one of these 3 say that “global standards is a best
option. But I doubt if such thing exists” (response from Founder and CEO,
i.e., R11). From the remaining responses, R5 says DK (don’t know).
a) What precautions can be taken to protect client data while inter-

facing between multiple cloud service providers?
Responses: Total Respondents to this question are 14. Given below are the
solution/practice/guideline mentioned against each respondent.
R2 Make sure that Vendor follow compliance best practices like ISO27k,
SOX, etc
R3 Encryption between Data and Other service provider . Network Secu-
rity and Firewall Blockage at and within Data center.
R4 Data Integration Data Encryption and security Data Authentication
Usage Policy
R5 DK
R6 DATA SECURITY IS CRITICAL. TYPICALLY I HAVE NOT SEEN

FOOL PROOF DATA SECURITY BETWEEN MULTIPLE SER-
VICE PROVIDERS. IT IS BEST TO RESTRICT DATA IN AN EX-
TENDED PRIVATE NETWORK OVER A SINGLE CLOUD PROVIDER
R7 Use proper authentication, security for each service provider.
R8 Have public cloud and check for sql injections and security measures
that each cloud vendor offers. Have continuous monitoring of data
slippage.
R9 Can’t comment.. have not thought about it.
R10 Again, if well-defined standards (similar to HIPAA for transfer of
healthcare information across service providers) exist, the client data
will be well-protected and strict rules will be defined for transfer be-
tween multiple service providers.
R11 High encryption Perimeter defense Access control Controlled ports
Isolation
R12 Do not trust any declarations of SaaS providers regarding security.
Protect your channel to SaaS from the integration hub (this is one
of the major roles of having such hub) with all security means your
corporate policies specify. If your applications are deployed in another
cloud, the communication channel with this one has to be equally
protected.
R13 Secure connections, multi-layer access controls and user education
R14 (1) TDE implementations are already in industry.
(2) 256 bit encryption
(3) Three phase encryption with hardware acceleration
(4) DLP mechanisms with secure replication
R15 Scalability,security and performance
Observations:
Most of the responses (6 out of 14) given to this question consist encryption
as the solution, such as: 256-bit encryption, data encryption, encryption
between data and other service provider, etc. Answers that need to be
noted are:
• “Data security is critical. Typically I have not seen fool proof data
security between multiple service providers. It is best to restrict data
in an extended private network over a single cloud provider” (R6, 2
years experience).
• “Do not trust any declarations of SaaS providers regarding security.

Protect your channel to SaaS from the integration hub (this is one
of the major roles of having such hub) with all security means your
corporate policies specify. If your applications are deployed in another
cloud, the communication channel with this one has to be equally
protected.” (R12, 7years experience).
As an answer to this question most of the respondents say that, there are
security measures such as encryption techniques or secure connections with
multi layered access control to be implements. From the above-mentioned
points, it makes clear that cloud computing is not secure and interconnec-
tion between cloud is to be taken extreme care.
Second question:
a) If service provider allows migration, what steps (methods used,

precautions taken or guidelines) can be followed to provide easy
migration without compromising security?
Responses:
Total Respondents to this question are 12. The responses of each respondent
in detail are as follows
R2 Not sure. No experience on this

R3 Data Center and Other service provider should be used the steps like
Network Security and Firewall Blockage at and within Data center.
R4 There are some set of rules to be followed prior to migration. both the
parties must agree on then, then only migration is possible
R5 DK
R7 Nothing have any thought
R8 (1) Have a sound schema - which should be in your full control.
(2) Don’t rely too much on services offered by vendor in core areas.
(3) Check the new cloud vendor offering with integrating your soft-
ware, with security offerings
R9 Cant comment as we are not using PaaS
R10 Transfer of security policies across multiple service providers will not
be possible until there is a common underlying framework that all
service providers implement. Since this does not exist, we are left to
individual expert consultants to advice on migration.
R11 Standard interfaces and data formats
R12 A workload that executes in one cloud provider can be uploaded to

another cloud provider. Some standardization efforts that support
this use case are Amazon Machine Image (AMI), Open Virtualization
Framework (OVF), and Virtual Hard Disk (VHD). Data that resides
in one cloud provider can be moved to another cloud provider. A stan-
dardization effort that supports this use case is Cloud Data Manage-
ment Interface (CDMI). In addition, even though SOAP and REST are
not data-specific standards, multiple cloud-storage providers support
data- and storage-management interfaces that use SOAP and REST.
A user who has established an identity with a cloud provider can use
the same identity with another cloud provider. Standardization efforts
that support this use case are Amazon Web Services Identity Access
Management (AWS IAM), OAuth, OpenID, and WS-Security.
R13 (1) Create robust clauses within the contracts - to ensure timely and
accurate extractions of data, followed by data erasal, upon termi-
nation of services
(2) Create maps between the source and target service providers - to
ensure seamless data migration
(3) Always keep periodic backups of essential/critical data, onsite.
Needless to say, most confidential information is better stored on-
premise
R14 This is already in place. MSFT to G to Amazon (tools exists for im-
port) Security for transfer of data/meta-data is done with encryption.
problems occur if domains do not allow rights keys.
Observations:
Three responder didnt answer this question i.e., people with 2.5years, 2years
and 1-year experience had no answer. Those who have answered have given
unique steps to be followed to handle the secure migration in a cloud com-
puting environment. Interesting responses to this question are as follows:
• “Transfer of security policies across multiple service providers will not

be possible until there is a common underlying framework that all
service providers implement. Since this does not exist, we are left to
individual expert consultants to advice on migration.” Which means
having a common set of rules followed by different service providers
could benefit migration of data (R10, 5years experience, founder and
CEO).
• “This is already in place. MSFT to G to Amazon (tools exists for im-
port) Security for transfer of data/meta-data is done with encryption.
Problems occur if domains do not allow rights keys” Which means that
there a migration assistance provided by service providers (response by

R14).
• The respondent R12, has listed out the frameworks, protocols and
standards that are helpful to look at when a user wishes to migrate his
system from one CSP to another. In this case his responses are more
specific to the Amazon service provider.
a) If service provider does not allow migration, what precautions

need to be followed to avoid data lock-in?
Responses:
The total respondents to this question are 10 and their answers as follows:
R2 Need to ensure that contracts are written in a way that we don’t get
into this sort of mess
R3 Smooth network traffic within and out of data center. You can used
agreement , RAA (Risk Analysis Acceptance) between vendor and
service provider during migration activity
R4 I am not able to comment on this topic
R5 DK
R8 Regular back up of data in your personal system space.
R10 We need to ensure that the application inherently provides the ability
to export data in some form that will make it easy to migrate to a
different provider.
R11 Store data internally, and use service provider only for processing.
R12 According to me the below measures should be taken to avoid vendor
lock-in in case a migration path does not exist :
(1) Data movement and encryption, both in transit and when it reaches
the target environment.
(2) Setting up networking to maintain certain relationships in the
source environment and preparing to connect into different net-
work options provided by the target environment.
(3) The application itself, which lives in an ecosystem surrounded by
tools and processes. When the application is moved to a target
cloud, you may have to re-architect it based on the components/re-
sources that the target cloud provides.
R13 (1) Find ways to replicate data that’s stored in the cloud
(2) Avoid storing sensitive data off-premise
R14 (1) read the fine print.

(2) Most service providers will give back the data as is
(3) Meta data management is usually a hurdle.
Observations:
To avoid this data lock-in due to no migration can be solved by following
precautions suggested by respondents:
(a) Ensure that the contract are written well

(b) Store data internally and use the cloud for processing
(c) Replicate data stored in the cloud and also avoid storing sensitive info
(d) Design an application on cloud interface that could give export user
data. This will solve the problem of exporting or fetching data manu-
ally.
Third Question:
a) What steps can be employed to provide compatibility between on-

premise authentication systems and cloud computing systems?
Responses: Total Respondents observed to this question in survey are 10.
Following are the responses mentioned against the respondents notation.
R2 For Cloud to be successfully vendors need to figure out how to inte-

grate all of these prevailing identify management systems. Can’t keep
reinventing stuff.
R3 Restricted Zone , Limited access in production zone
R5 DK
R8 Have an interface between your application and social media login’s ,
populate your DB with the same on regular basis and then allow users
to use the req. uid/passwd accordingly.
R10 Not sure.
R11 Similar software/hardware stacks Same patch levels on both Pre-
production validation environments
R12 I have not experienced this scenarios, but my suggestion would be to
adopt some in-house development of modules to help smooth interac-
tion between on-premise authentication systems and cloud computing
systems. Since organization confidential and user credential data flow
is involved via such interfaces, encryption would be recommended for
such communication channels.
R13 (1) Create an on-premise layer of authentication

(2) Explore security brokers to help integrate on-premise and off-
premise systems.
(3) Setup a local web server (with on-premise security) to serve up
data to the cloud users directly from the the local web server, so
there is a physical disconnect between the data in cloud and the
data on the local web server but to the user it looked like all the
data was in cloud.
R14 Firstly IDM is not enough. Identity is to be tagged across application,
users and hardware.. e.g. Mr. X can run Program Y on server Z and
W only... this is achieved with compatible signed hardware software
combos.
Observations:
There are 4 respondents who had no experience in this situation or had
no answer (7years, 5 years, 2.5 and 2 years). A respondent (person who is
part of 6 of the largest implementations in the industry. Across platforms,
engineering and hosting) says “Firstly IDM is not enough. Identity is to
be tagged across application, users and hardware.. e.g. Mr. X can run
Program Y on server Z and W only... This is achieved with compatible
signed hardware software combos.”
Another respondent says, “For Cloud to be successfully vendors need to
figure out how to integrate all of these prevailing identify management sys-
tems. Can’t keep reinventing stuff.” There were few opinions, which sug-
gested how to create compatibility between on-premise and cloud computing
systems.
a) What on-premise authentication system compatible with cloud

computing system do you use?
Responses: Total Respondents to this question are 10. Following are the
responses:
R2 LDAP Intergration
R3 It is very useful as every employee who is a part of CC they have to fol-
low the rules by cc as limited permission within and out of production
zone , never share your password with others.
R5 DK
R8 It’s always preferable to use SSO login, which is in your control.
R10 Haven’t used anything yet.
R11 o-auth
R12 SSO, LDAP, MS Windows authentication etc
R13 Oracle Identity and Access Mgmt solution
R14 Many.
Observations:
From the responses it can be identified that there are a few respondents
unaware of the IDM used. Others mention all most all the IDMs identi-
fied through literature review. Following table 5.1 shows which on-premise
authentication system compatible with which service provider in a cloud
computing system are used. For example, from the table 5.1, we can un-
derstand that Director for IT/SW Services who has experience with cloud
applications for a year says LDAP integration is compatible with google
cloud. The other respondent in this survey who is an associate director and
has been using Amazon cloud for 3 years says its preferable to use SSO (sin-
gle sign-on) as it is compatible with cloud applications. Similarly the other
responses have been tabulated to show which on-premise authentication
system is compatible with and from which service provider.
Table 5.1: Compatible on-premise authentication systems used with cloud
Position (experience Type of cloud Type of compatible on-

with cloud in years) used premise authentication
system used
Director IT/SW Services Google cloud LDAP integration
(1 year)
Associate director (3 Amazon cloud Preferable to use SSO
years)
Senior technology prac- Amazon cloud O-auth
tice manager (5 years)
Solution architect (7 Microsoft and SSO LDAP MS windows
years) Amazon cloud authentication etc.
Strategy and governance Amazon and Oracle identity and access
leader (8 years) salesforce.com management solution
Chief architect (part of 6 All the service Says many!
largest implementations providers
in the industry)
5.3 Summary
This section summarizes the observations of the survey based on the following.
1. Through this survey we have answers (guidelines) to the problems identified

in literature and presented in the survey questions:
a. Avoid customer lock-in.

b. Secured migration between two or more service providers (an example
is identified in the survey that shows Google to Amazon tools exist).
c. We have also a list of on-premise IDMs that are compatible with cloud
computing service providers.
2. Answers given to every question in the survey by respondents is different

and is a guideline or precautionary measure that users can follow to over-
come/avoid the problem mentioned.
3. Few users stress the use of global standards or common standards (to pro-
vide interoperability) and standard interfaces (for migration), as this did
not quite exist in the cloud computing scenarios.
4. Some of the answers mention that there is utmost care required when data
is migrated, interfaced or interconnected in cloud computing scenario. The
respondents mention different ways of encryption and security measures
that are to be implemented while transferring information from one cloud
to another.
5. From these surveys it can also be understood that framing strict standards
and having open source ways of designing CC applications is also stressed
for better security and compatibility of different cloud service providers.
Chapter 6
Discussion
Cloud computing is the most interesting technology among organizations and

most of these are organizations are concerned about its security. To identify CC
security concerns a literature review and snowball sampling were employed, which
revealed 43 challenges and 89 mitigation strategies. The identified challenges are
categorized and are explained in chapter 4. In addition to the challenges and
the mitigation strategies there are few guidelines and suggestions observed from
literature. The identified mitigation strategies and guidelines are discussed in
section 4.3. The identified challenges which can be mitigated using the observed
strategies are tabulated in section 4.4. This helps the reader to identify the
mitigation strategies that can be applied to effectively handle a challenge when
encountered during CC implementation. All this done to answer the research
question 1 and 2, which is to identify the security challenges and the mitigations.
The categorization of challenges into groups used in this report is done to simplify
explaining challenges. Grouping of challenges in this way was also used by other
authors (as been highlight in section 2.7), it was also mentioned that using another
type of categorization is suggested in further studies to avoid the missing of
security requirements.
Observations from the literature clearly state that most of the challenges iden-
tified through literature have at least one solution/guideline/practice/architec-
ture/model specified that can effectively handle the challenge. From the tables
and results mentioned in ‘section 4.4’ it can be observed that there are some
challenges which have no mitigation strategies. The challenges with no mitiga-
tion strategies are, ‘Packet sniffing’, ‘Sharing computing resources’, ‘Failure in
providing security’, ‘Incompatibility issue’, ‘Constant feature addition’, ‘Recov-
ery’, ‘Compliance’ and ‘Cost’. This study focuses on one challenge that has more
article mentioning the challenge and no mitigation strategies identified through
literature. The challenge chosen to be focused in this report is ‘Incompatibility’.
This is because there are 19 articles mentioning the challenge but no dependable
mitigation strategies identified.
Later, for this challenge identified (i.e., incompatibility), a survey method was
employed to gather expert opinion on mitigation strategies from practitioners.
To do this author needs to be aware of areas where this challenge can affect the
101
Chapter 6. Discussion 102
practitioners and then questionnaire for survey need to be framed. The insecure
areas of this challenge were identified from the literature review and are as follows:
1. Interoperation between different service providers,
2. Migration of data, application from one service provider to another,
3. Compatibility of traditional (on-premise) authentication methods with cloud

security systems.
From literature review there were very vague practices/suggestion/mitigation

strategies identified, which suggest using of open cloud manifesto standards.
These were not mentioned to be used by any company (observed from chapter
4). The survey questionnaire in this study was designed by taking into account
the above mentioned points and focused on gathering expert opinion from prac-
titioners. This answers the 3rd research question. The survey questionnaire is
presented in appendix B and the list of respondents is presented in table C.1
(due to confidentiality reasons detailed information of respondents can not be
presented).
The results of the surveys show that there is a need for security. The re-
spondents give guidelines, some precautions and their experiences as to how the
challenge situations mentioned in the questionnaire can be handled. From the
above, 3 insecure areas in cloud, there have been satisfactory responses from sur-
veys (observed from experienced people in cloud computing). The respondents
stress if global standards are available the incompatibility issue with CC (prob-
lems of interoperability between different service provider) can be avoided. In
addition to this it is also observed that, while interoperability being facilitated
between multiple service providers, utmost care is required to be taken and chan-
nels through which data is being communicated need to be secured. Failing to
do so can be risky and also information leakage (number of preventive measures
are given to facilitate secure interoperability between cloud providers). From the
surveys list of all compatible on-premise authentication systems that can be used
with cloud computing and what steps should be followed to provide secure migra-
tion between two or more service providers (to avoid data lock-in situations) are
observed. Detailed survey responses to these questions are discussed in chapter
5.
6.1 Validity threats

The following are the validity threats that can be identified in this thesis.
• There might be a chance of missing papers when selecting initial papers for
snowball sampling. To avoid this authors of initial papers were contacted
and requested to suggest papers that might be helpful when included in
our study. In addition to this the author performed a literature review in
addition to snowball sampling to strengthen the results of this study.
• Since there are numerous challenges identified from literature review and if
they are not properly recorded there are chances of missing some important
challenges or solutions. To prevent this from happening there was a data
extraction form maintained which helped in tracking challenges identified.
• Too many challenges listed in an unorganized way, can confuse the reader.
To avoid this, challenges are categorized under 8 sections, some of them
might find this categorization unclear and hard to understand what they
constitute. To avoid this confusion, each categorization is explained clearly
in terms, which were collectively grouped under them.
• The challenges were categorized into different sections and few challenges
with similar threat and risk were grouped under a challenge. Based on
this categorization the identified mitigation strategies were also mentioned.
Possibility of grouping irrelevant challenges and mitigation strategies can
happen, which is avoided by giving a detailed explanation of each challenge
and their mitigation strategies. Doing this will remove confusion among the
readers.
• Collecting information from a large group of practitioners could have biased
results. There might be chances for people replying even when they are not
related to cloud computing. To avoid these practitioners who have high
experience in this field were contacted after searching in different forums
and companies that implement cloud.
• Since the surveys were conducted through a survey link, there is possibility
that respondents find it hard to understand the purpose of the survey. To
avoid this issue each respondent sent a startup mail in addition to the
introduction of purpose mentioned before presenting questions in the survey
link.
• There is a possibility for responders finding it difficult to understand the
terms used in interviews. To avoid this issue, interviews were conducted
only with those persons who have worked in this area. They were also
provided with an e-mail id, through which respondents could contact the
authors and clear their doubts.
• There is a threat that readers might consider these results are generalized,
to avoid this problem it is stated in the report that these generalized results
are obtained from expert opinion or from literature and can differ with the
change from organization to organization.
Chapter 7
Conclusions and Future Work
7.1 Conclusions
This study collectively describes cloud computing security challenges in general
and describes the mitigation practices that have been proposed to handle the
identified challenges. But there are still some challenges with no mitigation strate-
gies, which might stand as a risk and a concern for some enthusiastic CC lovers.
Through this study the author tried to focus on one such challenge ‘incompati-
bility’ and find mitigation practices from CC practitioners.
The conclusion shall be divided into 3 subsections and inferences obtained
from each research question is explained below:
RQ1 & RQ2: After an exhaustive search (using literature review and snowball
sampling) on electronic databases, 43 challenges and 89 mitigation strategies
were identified in cloud computing security. In addition to these mitigation
strategies there are also some guidelines been identified.
RQ3: The guidelines and mitigation strategies identified through initial study are
not available for all the challenges identified. There are some challenges with
no mitigation strategies and from them a challenge that is mentioned by
multiple articles is selected. For this challenge (‘incompatibility’) identified
from literature that has no mitigation strategies, solutions/practices/miti-
gation strategies are found from cloud computing practitioners.
RQ4: Based on the survey method applied to identify the practices to mitigate
this ‘incompatibility’ challenge there are some conclusions made and as
follow:
1. There are ways to provide interoperability between multiple cloud ser-

vice providers (data/work load/instances/application), but there is no
proper standard or standard interface observed by cloud users. Each
of them have designed a series of steps through which this is intercon-
nectivity is established. It can also be observed that this individually
designed steps by users could have flaws as they are not standardized.
105
Chapter 7. Conclusions and Future Work 106
Hence, the suggested methods need to be deployed with extreme cau-

tion to prevent security risks (can be observed by the responses from
the survey that suggest the usage of multiple encryption techniques at
various levels to prevent data leakage)
2. For migration of data, an issue pointed in survey, the responses state
that there are some standards user can look at while making a deci-
sion to migrate data/application to another CSP (but these are given
in specific to Amazon services). These standards not all cloud users
are familiar, which is because there is no common framework that is
existing with the CSP to provide migration. In addition to this, in
case migration is not allowed by the service provider, the best way
to stay secure and safe from data lock-in is by not uploading sensi-
tive information to the cloud and make sure that the contract is well
written.
3. For the on-premise authentication (IDM’s) being compatible with cloud
security, the conclusions can be that there are no proper seamless inte-
gration strategies existent and most of the cloud user have to depend
on vendors. In addition to this, there are some set of on-premise au-
thentication strategies need to be identified which can be integrated
with cloud service providers that are mentioned.
7.2 Future work

There are very few studies that discuss all these 43 challenges and 89 solutions.
Among these challenges, some do not have solutions or clear description. These
need to be evaluated further to understand their impact on cloud computing and
then discarded or practices should be designed.
However, most important future work identifies here is that there are con-
crete standards for cloud computing security still missing. There are some open
cloud manifesto standards and few efforts made by the cloud security alliance
to standardize the process in the cloud. The cloud vendors and users don’t en-
courage the usage of these standards as they are restrictive. In addition to this
the cloud computing with such great offering such as storage, infrastructure and
application designing capabilities on the go to the IT industry still fail to have
proper standards for interoperability with other cloud service providers. This
failure to provide concrete security standards, common underlying framework for
data migration and global standards for cloud interoperability, make the leading
technology “the cloud computing” still a vulnerable option for aspiring users.
References
[1] Top threats to cloud computing : Cloud security alliance.

https://cloudsecurityalliance.org/research/top-threats/. 14
[2] Cloud cube model: Selecting cloud formations for secure collaboration,
April 2009. 25
[3] Security in the cloud. Clavister White Paper, 2010. cited By (since 1996)
1. 8, 25, 30, 36, 43, 54, 60, 63, 66, 70
[4] Imad M. Abbadi and Cornelius Namiluko. Dynamics of trust in Clouds—

Challenges and research agenda. In Internet Technology and Secured Trans-
actions (ICITST), 2011 International Conference for, pages 110–115, 2011.
46
[5] Hussain Al-Aqrabi, Lu Liu, Jie Xu, Richard Hill, Nick Antonopoulos, and
Yongzhao Zhan. Investigation of IT security and compliance challenges in
security-as-a-service for cloud computing. In Object/Component/Service-
Oriented Real-Time Distributed Computing Workshops (ISORCW), 2012
15th IEEE International Symposium on, pages 124–129, 2012. 33, 37, 40,
44, 49, 75
[6] M. Al Morsy, J. Grundy, and I. Müller. An analysis of the cloud computing

security problem. In the proc. of the 2010 Asia Pacific Cloud Workshop,
Colocated with APSEC2010, Australia, 2010. 42, 43, 45, 48, 68, 69
[7] Aiiad Ahmad Albeshri and William Caelli. Mutual protection in a cloud
computing environment. In IEEE 12th International Conference on High
Performance Computing and Communications (HPCC 2010), pages 641–
646, 2010. 11, 27, 33, 37, 46, 52, 62, 75
[8] Abdulrahman Almutairi, Muhammad Sarfraz, Saleh Basalamah, Walid

Aref, and Arif Ghafoor. A distributed access control architecture for cloud
computing. Software, IEEE, 29(2):36–44, 2012. 51, 72
[9] William; Athley Ambrose. Cloud Computing : Security Risks, SLA, and
Trust. 2010. With Cloud Computing becoming a popular term on the In-
formation Technology (IT) market, security and accountability has become
107
References 108
important issues to highlight. In our research we review these concept ...

10, 15, 18
[10] T. Andrei and R. Jain. Cloud computing challenges and related security
issues. A Survey Paper. DOI= http://www. cse. wustl. edu/ jain/cse571-
09/ftp/cloud. pdf. 25, 26, 32
[11] Gabriel Antoniu. Autonomic cloud storage: challenges at stake. In Com-

plex, Intelligent and Software Intensive Systems (CISIS), 2010 Interna-
tional Conference on, pages 481–481, 2010. 29, 35
[12] M. Armbrust, A. Fox, R. Griffith, A.D. Joseph, R.H. Katz, A. Konwinski,

G. Lee, D.A. Patterson, A. Rabkin, I. Stoica, et al. Above the clouds:
A berkeley view of cloud computing. EECS Department, University of
California, Berkeley, Tech. Rep. UCB/EECS-2009-28, 2009. 33, 34, 52, 53,
55, 62, 66, 67, 69, 70, 71
[13] Michael Armbrust, Armando Fox, Rean Griffith, Anthony D Joseph, Randy
Katz, Andy Konwinski, Gunho Lee, David Patterson, Ariel Rabkin, Ion
Stoica, and Matei Zaharia. A view of cloud computing. Communications
of the ACM, 53:50–58, April 2010. ACM ID: 1721672. 33, 34, 52, 53, 55,
62, 66, 67, 69, 70, 71
[14] Junaid Arshad, Paul Townend, and Jie Xu. A novel intrusion severity
analysis approach for clouds. Future Generation Computer Systems, 2011.
36, 72
[15] Akhil Behl. Emerging security challenges in cloud computing: An insight

to cloud security challenges and their mitigation. In Information and Com-
munication Technologies (WICT), 2011 World Congress on, pages 217–222,
2011. 32, 37, 41, 74, 78
[16] D. Bernstein, E. Ludvigson, K. Sankar, S. Diamond, and M. Morrow.

Blueprint for the intercloud-Protocols and formats for cloud computing in-
teroperability. In 2009 Fourth International Conference on Internet and
Web Applications and Services, pages 328–336, 2009. 54
[17] Aashish Bhardwaj and Vikas Kumar. Cloud security assessment and iden-
tity management. In Computer and Information Technology (ICCIT), 2011
14th International Conference on, pages 387–392, 2011. 24, 33, 37, 40, 49,
58, 72
[18] W. Bin, H.H. Yuan, L.X. Xi, and X.J. Min. Open identity management
framework for SaaS ecosystem. In 2009 IEEE International Conference on
e-Business Engineering, pages 512–517, 2009. 48
References 109
[19] K. Birman, G. Chockler, and R. van Renesse. Toward a cloud computing

research agenda. SIGACT News, 40(2):68–80, 2009. 28
[20] M. Blaze, J. Feigenbaum, J. Ioannidis, and A. Keromytis. The role of trust

management in distributed systems security. Secure Internet Programming,
pages 185–210, 1999. 50
[21] M. Blaze, S. Kannan, I. Lee, O. Sokolsky, J.M. Smith, A.D. Keromytis, and
W. Lee. Dynamic trust management. Computer, 42(2):44–52, 2009. 70
[22] J. Brodkin. Gartner: Seven cloud-computing security risks. Infoworld,

pages 1–3, 2008. 27, 30, 60, 63, 66, 67
[23] Yuriy Brun and Nenad Medvidovic. Keeping data private while computing
in the cloud. In Cloud Computing (CLOUD), 2012 IEEE 5th International
Conference on, pages 285–294, 2012. 57, 59, 75
[24] R. Buyya, C.S. Yeo, and S. Venugopal. Market-oriented cloud computing:

Vision, hype, and reality for delivering it services as computing utilities.
In The 10th IEEE international conference on high performance computing
and communications, pages 5–13, 2008. 38, 54
[25] Rajkumar Buyya, Chee Shin Yeo, Srikumar Venugopal, James Broberg, and
Ivona Brandic. Cloud computing and emerging IT platforms: Vision, hype,
and reality for delivering computing as the 5th utility. Future Generation
Computer Systems, 25(6):599–616, June 2009. 42
[26] C. Cachin, I. Keidar, and A. Shraer. Trusting the cloud. ACM SIGACT
News, 40(2):81–86, 2009. 25, 26, 34, 35, 71
[27] Jose M. Alcaraz Calero, Nigel Edwards, Johannes Kirschnick, Lawrence

Wilcock, and Mike Wray. Toward a multi-tenancy authorization system for
cloud services. Security & Privacy, IEEE, 8(6):48–55, 2010. 78
[28] D. Catteddu. Cloud computing: benefits, risks and recommendations for

information security. Web Application Security, pages 17–17, 2010. 25, 26,
63
[29] A. Cavoukian. Privacy in the clouds. Identity in the Information Society,

1(1):89–108, 2008. 72
[30] Stuart Charters and Barbara Kitchenham. Guidelines for performing sys-
tematic literature reviews in software engineering. (EBSE 2007-001), 2007.
19
References 110
[31] Ainul Che Fauzi, A. Noraziah, Tutut Herawan, and Noriyani Mohd. Zin.
On cloud computing security issues. Intelligent Information and Database
Systems, pages 560–569, 2012. 28, 35, 38, 39, 52, 66
[32] Jianyong Chen, Yang Wang, and Xiaomin Wang. On-demand security ar-
chitecture for cloud computing. Computer, 45(7):73–78, 2012. 38, 39, 46,
50, 52, 74, 75
[33] Lanxiang Chen and Gongde Guo. An efficient remote data possession check-
ing in cloud storage. International Journal of Digital Content Technology
and its Applications, 5(4):43–50, 2011. 31, 34, 73
[34] Y. Chen, V. Paxson, and R.H. Katz. What’s new about cloud computing
security. University of California, Berkeley Report No. UCB/EECS-2010-5
January, 20(2010):2010–5, 2010. 40, 44
[35] R. Chow, P. Golle, M. Jakobsson, E. Shi, J. Staddon, R. Masuoka, and
J. Molina. Controlling data in the cloud: outsourcing computation without
outsourcing control. In Proceedings of the 2009 ACM workshop on Cloud
computing security, pages 85–90, 2009. 34, 52, 70
[36] John W. Creswell. Research design: Qualitative, quantitative, and mixed
methods approaches. Sage Publications, Incorporated, 2008. 20, 86
[37] William L Curlette and Kathleen Silva Cannella. Research in Nursing &
Health, 8(3):293–301, September 1985. 24
[38] W. Dawoud, I. Takouna, and C. Meinel. Infrastructure as a service security:
Challenges and solutions. In 2010 The 7th International Conference on
Informatics and Systems (INFOS), pages 1–8. IEEE, March 2010. 32, 35,
37, 41, 42, 43, 44, 46, 50
[39] T. Dillon, Chen Wu, and E. Chang. Cloud computing: Issues and chal-
lenges. In 2010 24th IEEE International Conference on Advanced Informa-
tion Networking and Applications (AINA), pages 27 –33, April 2010. 3, 4,
8, 9, 10, 11
[40] Mary Dixon-Woods, Shona Agarwal, David Jones, Bridget Young, and Alex
Sutton. Synthesising qualitative and quantitative evidence: a review of
possible methods. Journal of Health Services Research & Policy, 10(1):45
–53B, January 2005. 24
[41] Fred Douglis. Staring at clouds. Internet Computing, IEEE, 13(3):4–6,
2009. 55
[42] R. Dubey, M.A. Jamshed, X. Wang, and R.K. Batalla. Addressing security
issues in cloud computing, 2012. 59, 71
References 111
[43] Adrian J. Duncan, Sadie Creese, and Michael Goldsmith. Insider attacks
in cloud computing. In Trust, Security and Privacy in Computing and
Communications (TrustCom), 2012 IEEE 11th International Conference
on, pages 857–862, 2012. 50
[44] BY Dustin oWens. Securing elasticity in the cloud. Communications of the

ACM, 53(6), 2010. 41, 64, 65
[45] Victor Echeverria, Lorie M. Liebrock, and Dongwan Shin. Permission man-
agement system: Permission as a service in cloud computing. In Com-
puter Software and Applications Conference Workshops (COMPSACW),
2010 IEEE 34th Annual, pages 371–375, 2010. 30, 34, 76
[46] L. Ertaul, S. Singhal, and G. Saldamli. Security challenges in cloud com-

puting. California State University, East Bay. Academic paper http://www.
mcs. csueastbay. edu/ lertaul/Cloudpdf, 2009. 12, 15, 18, 27, 28, 33, 38, 39,
40, 47, 54, 56, 64, 65, 66, 67, 68
[47] D.M. Eyers, R. Routray, R. Zhang, D. Willcocks, and P. Pietzuch. Towards

a middleware for configuring large-scale storage infrastructures. In Proceed-
ings of the 7th International Workshop on Middleware for Grids, Clouds
and e-Science, page 3, 2009. 72
[48] Sascha Fahl, Marian Harbach, Thomas Muders, and Matthew Smith. Con-
fidentiality as a Service–Usable security for the cloud. In Trust, Security
and Privacy in Computing and Communications (TrustCom), 2012 IEEE
11th International Conference on, pages 153–162, 2012. 57, 74
[49] F. Farahmand. Risk perception and trust in cloud. Information Systems

Control Journal, (4):8 pp., 2010. 45, 58, 76
[50] Christoph Fehling, Thilo Ewald, Frank Leymann, Michael Pauly,

J. Rutschlin, and David Schumm. Capturing cloud computing knowledge
and experience in patterns. In Cloud Computing (CLOUD), 2012 IEEE 5th
International Conference on, pages 726–733, 2012. 40, 41, 42, 45, 55, 63
[51] Jun Feng, Yu Chen, Wei-Shinn Ku, and Pu Liu. Analysis of integrity vul-
nerabilities and a non-repudiation protocol for cloud data storage platforms.
In The 2nd International Workshop on Security in Cloud Computing (SCC
2010), in Conjunction with ICPP, 2010. 18, 36, 37, 61, 73
[52] Arlene G. Fink. Conducting Research Literature Reviews: From Paper to

the Internet. Sage Publications, Inc, first edition edition, April 1998. 19
References 112
[53] Baruch Fischhoff, Paul Slovic, Sarah Lichtenstein, Stephen Read, and Bar-
bara Combs. How safe is safe enough? a psychometric study of attitudes to-
wards technological risks and benefits. Policy sciences, 9(2):127–152, 1978.
76
[54] Teodor-Florin Fortis, Victor Ion Munteanu, and Viorel Negru. Steps to-
wards cloud governance. a survey. In Information Technology Interfaces
(ITI), Proceedings of the ITI 2012 34th International Conference on, pages
29–34, 2012. 32, 33, 57, 79
[55] I Foster, Yong Zhao, I Raicu, and S Lu. Cloud computing and grid com-
puting 360-degree compared. In Grid Computing Environments Workshop,
2008. GCE ’08, pages 1–10. IEEE, November 2008. 3, 25, 27, 29, 34, 52,
54, 59, 60, 63
[56] Volker Fusenig and Ayush Sharma. Security architecture for cloud net-
working. In Computing, Networking and Communications (ICNC), 2012
International Conference on, pages 45–49, 2012. 43, 77
[57] Gebeyehu Belay Gebremeskel, Chengliang Wang, and Zhongshi He. The
paradigm integration of computational intelligence performance in cloud
computing towards data security. In Information and Computing Science
(ICIC), 2012 Fifth International Conference on, pages 19–22, 2012. 25, 76
[58] R. Gellman. Privacy in the clouds: Risks to privacy and confidentiality

from cloud computing. In World privacy forum, pages 1–26, 2009. 25
[59] E.J. Goh, H. Shacham, N. Modadugu, and D. Boneh. SiRiUS: securing

remote untrusted storage. In Proc. NDSS, volume 3, 2003. 71
[60] L. A Goodman. Snowball sampling. The Annals of Mathematical Statistics,

32(1):148–170, 1961. 19
[61] A. Gopalakrishnan. Cloud computing identity management. SETLabs

Briefings, 7(7):45–54, 2009. 55, 56, 70
[62] B. Grobauer, T. Walloschek, and E. Stöcker. Understanding cloud comput-

ing vulnerabilities. IEEE Security and Privacy, pages 50–57, 2010. 50
[63] R.L. Grossman. The case for cloud computing. it Professional, 11(2):23–27,
2009. 25, 56
[64] Qiang Guo, Dawei Sun, Guiran Chang, Lina Sun, and Xingwei Wang. Mod-
eling and evaluation of trust in cloud computing environments. In Advanced
Computer Control (ICACC), 2011 3rd International Conference on, pages
112–116, 2011. 46, 75
References 113
[65] Mohamed Hamdi. Security of cloud computing, storage, and networking.

In Collaboration Technologies and Systems (CTS), 2012 International Con-
ference on, pages 1–5, 2012. 5, 32, 37, 38, 45, 52, 57, 63, 64, 72
[66] Kevin Hamlen, Peng Liu, Murat Kantarcioglu, Bhavani Thuraisingham,
and Ting Yu. Identity management for cloud computing: developments
and directions. In Proceedings of the Seventh Annual Workshop on Cyber
Security and Information Intelligence Research, page 32, 2011. 50, 69
[67] A. Hammami, N. Simoni, and R. Salman. Ubiquity and QoS for cloud se-
curity. In 2012 41st International Conference on Parallel Processing Work-
shops (ICPPW), pages 277 –278, September 2012. 38, 39, 41, 46, 57, 76
[68] Shuai Han and Jianchuan Xing. Ensuring data storage security through a
novel third party auditor scheme in cloud computing. In Cloud Computing
and Intelligence Systems (CCIS), 2011 IEEE International Conference on,
pages 264–268, 2011. 34, 74
[69] S. Hanna. Cloud computing: finding the silver lining. Juniper Networks,
2009. 33
[70] Danny Harnik, Elliot K. Kolodner, Shahar Ronen, Julian Satran, Alexandra
Shulman-Peleg, and Sivan Tal. Secure access mechanism for cloud storage.
Scalable Computing: Practice and Experience, 12(3), 2011. 25, 37, 53, 77
[71] B. Hay, K. Nance, and M. Bishop. Storm clouds rising: Security challenges
for IaaS cloud computing. In 2011 44th Hawaii International Conference
on System Sciences (HICSS), pages 1–7. IEEE, January 2011. 40, 60, 78
[72] J. Heiser and M. Nicolett. Assessing the security risks of cloud computing.
Gartner Report, 2008. 27, 29, 52, 60
[73] Ghasem Heyrani-Nobari, Omar Boucelma, and Stéphane Bressan. Privacy
and anonymization as a service: PASS. In Hiroyuki Kitagawa, Yoshiharu
Ishikawa, Qing Li, and Chiemi Watanabe, editors, Database Systems for
Advanced Applications, volume 5982, pages 392–395. Springer Berlin Hei-
delberg, Berlin, Heidelberg, 2010. 9, 43, 44, 70
[74] Fei Hu, Meikang Qiu, Jiayin Li, Travis Grant, Drew Taylor, Seth McCaleb,
Lee Butler, and Richard Hamner. A review on cloud computing: Design
challenges in architecture and security. Journal of Computing and Infor-
mation Technology, 19(1):25–55, 2011. 43, 57, 72
[75] Iliana Iankoulova and Maya Daneva. Cloud computing security require-
ments: A systematic review. In Research Challenges in Information Science
(RCIS), 2012 Sixth International Conference on, pages 1–7, 2012. 3, 14,
16, 28, 37, 43, 44, 48, 58, 66, 73, 74
References 114
[76] P.T. Jaeger, J. Lin, and J.M. Grimes. Cloud computing and information
policy: Computing in a policy cloud? Journal of Information Technology
& Politics, 5(3):269–283, 2008. 30
[77] T. Jaeger and J. Schiffman. Outlook: Cloudy with a chance of security
challenges and improvements. Security & Privacy, IEEE, 8(1):77–80, 2010.
9, 46
[78] W. Jansen and T. Grance. Guidelines on security and privacy in public
cloud computing. NIST Draft Special Publication, pages 800–144, 2011. 26,
33, 43, 46, 49, 50, 52, 61, 63, 65, 69
[79] M. Jensen and J. Schwenk. The accountability problem of flooding at-
tacks in service-oriented architectures. In 2009 International Conference
on Availability, Reliability and Security, pages 25–32, 2009. 39
[80] M. Jensen, J. Schwenk, N. Gruschka, and L.L. Iacono. On technical security
issues in cloud computing. In 2009 IEEE International Conference on Cloud
Computing, pages 109–116, 2009. 15, 38, 39
[81] Zhang Jianhong and Chen Hua. Secuirty storage in the cloud comput-
ing: A RSA-based assumption data integrity check without original data.
In Educational and Information Technology (ICEIT), 2010 International
Conference on, volume 2, pages V2–143, 2010. 28, 71, 74, 77
[82] D. Johnson. Computing in the clouds. Learning & Leading with Technology,
page 10, 2009. 41
[83] Bansidhar Joshi, A. Santhana Vijayan, and Bineet Kumar Joshi. Securing
cloud computing environment against DDoS attacks. In Computer Com-
munication and Informatics (ICCCI), 2012 International Conference on,
pages 1–5, 2012. 39, 78
[84] James B. D. Joshi, Walid G. Aref, Arif Ghafoor, and Eugene H. Spafford.
Security models for web-based applications. Commun. ACM, 44(2):38–44,
February 2001. 41, 46
[85] Shailza Kamal and Rajpreet Kaur. Cloud computing security issue: Survey.
AIP Conference Proceedings, 1414(1):149–153, December 2011. 11, 25, 27,
29, 37, 52, 59, 60, 63, 69
[86] B.R. Kandukuri, VR Paturi, and A. Rakshit. Cloud security issues. In Ser-
vices Computing, 2009. SCC’09. IEEE International Conference on, pages
517–520, 2009. 7, 27, 30, 60, 66
[87] L.M. Kaufman. Data security in the world of cloud computing. IEEE
Security & Privacy, pages 61–64, 2009. 11, 47
References 115
[88] L.M. Kaufman. Can a trusted environment provide security? Security &
Privacy, IEEE, 8(1):50–52, 2010. 10, 71
[89] Khaled M. Khan and Qutaibah Malluhi. Establishing trust in cloud com-
puting. IT professional, 12(5):20–27, 2010. 46, 79
[90] Md Tanzim Khorshed, A. B. M. Ali, and Saleh A. Wasimi. A survey on

gaps, threat remediation challenges and some thoughts for proactive attack
detection in cloud computing. Future Generation Computer Systems, 2012.
4, 13, 15, 16, 31, 33, 36, 40, 41, 47, 49, 51, 57, 58, 73
[91] Michael Kretzschmar, Mario Golling, and Sebastian Hanigk. Security man-
agement areas in the inter-cloud. In Cloud Computing (CLOUD), 2011
IEEE International Conference on, pages 762–763, 2011. 45, 50, 54, 62
[92] G. Kulkarni, J. Gambhir, T. Patil, and A. Dongare. A security aspects in

cloud computing. In 2012 IEEE 3rd International Conference on Software
Engineering and Service Science (ICSESS), pages 547 –550, June 2012. 25,
33, 34, 37, 38, 47, 57, 63, 66
[93] Rikard Larsson. Case survey methodology: Quantitative analysis of

patterns across case studies. The Academy of Management Journal,
36(6):1515–1546, December 1993. ArticleType: research-article / Full pub-
lication date: Dec., 1993 / Copyright © 1993 Academy of Management.
24
[94] D. Lekkas. Establishing and managing trust within the public key infras-
tructure. Computer Communications, 26(16):1815–1825, 2003. 69
[95] Jin Li, Gansen Zhao, Xiaofeng Chen, Dongqing Xie, Chunming Rong, Wen-
jun Li, Lianzhang Tang, and Yong Tang. Fine-grained data access control
systems with user accountability in cloud computing. In Cloud Comput-
ing Technology and Science (CloudCom), 2010 IEEE Second International
Conference on, pages 89–96, 2010. 29, 34, 74
[96] G. Lin, G. Dasmalchi, and J. Zhu. Cloud computing and IT as a service:

opportunities and challenges. In Web Services, 2008. ICWS’08. IEEE In-
ternational Conference on, pages 5–5, 2008. 10
[97] Feng Liu, Weiping Guo, Zhi Qiang Zhao, and Wu Chou. SaaS integra-
tion for software cloud. In Cloud Computing (CLOUD), 2010 IEEE 3rd
[98] Wenjun Luo and Guojing Bai. Ensuring the data integrity in cloud data
storage. In Cloud Computing and Intelligence Systems (CCIS), 2011 IEEE
References 116
[99] Wenjun Luo and Guojing Bai. Multi-copy privacy-preserving verification

for cloud computing. International Journal of Advancements in Computing
Technology, 3(9):9–16, 2011. 28, 31, 75
[100] Xiaoqi Ma. Security concerns in cloud computing. In Computational and

Information Sciences (ICCIS), 2012 Fourth International Conference on,
pages 1069–1072, 2012. 11, 27, 28, 31, 33, 34, 39, 52, 66, 76
[101] F. Maggi and S. Zanero. Rethinking security in a cloudy world. Techni-

cal report, Technical report, Dipartimento di Elettronica e Informazione,
Politecnico di Milano, 2010. 12
[102] U. Maheshwari, R. Vingralek, and W. Shapiro. How to build a trusted

database system on untrusted storage. In Proceedings of the 4th conference
on Symposium on Operating System Design & Implementation-Volume 4,
pages 10–10, 2000. 71
[103] Patricia Yancey Martin and Barry A. Turner. Grounded theory and orga-
nizational research. The Journal of Applied Behavioral Science, 22(2):141
–157, April 1986. 24
[104] Tim Mather, Subra Kumaraswamy, and Shahed Latif. Cloud Security and
Privacy: An Enterprise Perspective on Risks and Compliance. O’Reilly
Media, 1 edition, October 2009. 48, 55, 57, 62, 65
[105] Eystein Mathisen. Security challenges and solutions in cloud computing. In

Digital Ecosystems and Technologies Conference (DEST), 2011 Proceedings
of the 5th IEEE International Conference on, pages 208–212, 2011. 15, 25,
27, 38, 41, 44, 48, 49, 52, 54, 59, 64, 71, 79
[106] Peter Mell and Tim Grance. The NIST definition of cloud computing.
National Institute of Standards and Technology, 53(6):50, 2009. 7, 9
[107] Parastoo Mohagheghi and Thor Sæther. Software engineering challenges

for migration to the service cloud paradigm. 25
[108] Eman M. Mohamed, Hatem S. Abdelkader, and S. El-Etriby. Enhanced

data security model for cloud computing. In Informatics and Systems (IN-
FOS), 2012 8th International Conference on, pages CC–12, 2012. 34
[109] M.B. Mollah, K.R. Islam, and S.S. Islam. Next generation of computing
through cloud computing technology. In 2012 25th IEEE Canadian Confer-
ence on Electrical Computer Engineering (CCECE), pages 1 –6, May 2012.
3, 25, 27, 29, 34, 41, 49, 54, 56, 57, 59, 63
References 117
[110] Aryan Taheri Monfared and Martin Gilje Jaatun. Monitoring intrusions
and security breaches in highly distributed cloud environments. In Cloud
Computing Technology and Science (CloudCom), 2011 IEEE Third Inter-
national Conference on, pages 772–777, 2011. 27, 33, 34, 37, 40, 46, 62
[111] M.C. Mont, S. Pearson, and P. Bramhall. Towards accountable management

of identity and privacy: Sticky policies and enforceable tracing services. In
Database and Expert Systems Applications, 2003. Proceedings. 14th Inter-
national Workshop on, pages 377–382, 2003. 57
[112] J. Morin, Jocelyn Aubert, and Benjamin Gateau. Towards cloud computing
SLA risk management: Issues and challenges. In System Science (HICSS),
2012 45th Hawaii International Conference on, pages 5509–5514, 2012. 48
[113] Antonio Muñoz, Javier Gonzalez, and Antonio Maña. A performance-

oriented monitoring system for security properties in cloud computing ap-
plications. The Computer Journal, 55(8):979–994, 2012. 72
[114] George W. Noblit and R. Dwight Hare. Meta-ethnography: synthesizing

qualitative studies. SAGE, February 1988. 24
[115] Brian O’Flaherty and Jason Whalley. Qualitative analysis software applied
to IS research - developing a coding strategy. ECIS 2004 Proceedings, Jan-
uary 2004. 23, 24
[116] D. Oppenheimer. The importance of understanding distributed system con-

figuration. In Proceedings of the 2003 Conference on Human Factors in
Computer Systems workshop, 2003. 52, 60
[117] S. Pearson. Taking account of privacy when designing cloud computing ser-
vices. In Proceedings of the 2009 ICSE Workshop on Software Engineering
Challenges of Cloud Computing, pages 44–52, 2009. 58
[118] Sanjeev Pippal, Vishu Sharma, Shakti Mishra, and D. S. Kushwaha. Secure
and efficient multitenant database for an ad hoc cloud. In Securing Services
on the Cloud (IWSSC), 2011 1st International Workshop on, pages 46–50,
2011. 77
[119] Kresimir Popovic and Zeljko Hocenski. Cloud computing security issues
and challenges. In MIPRO, 2010 Proceedings of the 33rd International
Convention, pages 344–349, 2010. 4, 8, 12, 26, 27, 32, 33, 34, 40, 48, 58
[120] B. Priyadharshini and P. Parvathi. Data integrity in cloud storage. In

Advances in Engineering, Science and Management (ICAESM), 2012 In-
ternational Conference on, pages 261–265, 2012. 28
References 118
[121] L. Qiu, Y. Zhang, F. Wang, M. Kyung, and H.R. Mahajan. Trusted com-
puter system evaluation criteria. In National Computer Security Center,
1985. 29
[122] Syed Mujib Rahaman and Mohammad Farhatullah. PccP: a model for pre-
serving cloud computing privacy. In Data Science & Engineering (ICDSE),
2012 International Conference on, pages 166–170, 2012. 59, 75
[123] S. Ramgovind, Mariki M. Eloff, and E. Smith. The management of security

in cloud computing. In Information Security for South Africa (ISSA), 2010,
pages 1–7, 2010. 8, 10, 27, 29, 35, 38, 40, 42, 46, 51, 52, 54, 57, 59, 60, 63,
67, 69
[124] R. Ranchal, B. Bhargava, LB Othmane, L. Lilien, A. Kim, M. Kang, and

M. Linderman. Protection of identity information in cloud computing with-
out trusted third party. In Reliable Distributed Systems, 2010 29th IEEE
Symposium on, pages 368–372, 2010. 48
[125] G. Reese. Cloud application architectures. O’Reilly Media, Inc., 2009. 32
[126] Kui Ren, Cong Wang, and Qian Wang. Security challenges for the public
cloud. Internet Computing, IEEE, 16(1):69–73, 2012. 27, 28, 32, 33, 35, 42,
44, 48, 57, 62
[127] B. P Rimal, Eunmi Choi, and I. Lumb. A taxonomy and survey of cloud
computing systems. In Fifth International Joint Conference on INC, IMS
and IDC, 2009. NCM ’09, pages 44–51. IEEE, August 2009. 9, 10, 47, 69
[128] T. Ristenpart, E. Tromer, H. Shacham, and S. Savage. Hey, you, get off of
my cloud: exploring information leakage in third-party compute clouds. In
Proceedings of the 16th ACM conference on Computer and communications
security, pages 199–212, 2009. 44
[129] Lois A. Ritter and Valerie M. Sue. Systematic planning for using an online
survey. New Directions for Evaluation, 2007(115):15–22, 2007. 87
[130] John Rittinghouse and James Ransome. Cloud Computing: Implementa-

tion, Management, and Security. CRC Press, 1 edition, August 2009. 8, 9,
10, 11, 28, 40, 54
[131] L. Rodero-Merino, L.M. Vaquero, V. Gil, F. Galán, J. Fontán, R.S. Mon-

tero, and I.M. Llorente. From infrastructure delivery to service management
in clouds. Future Generation Computer Systems, 26(8):1226–1240, 2010. 56
[132] J. Ruiter and M. Warnier. Privacy regulations for cloud computing. 34, 59,
65
References 119
[133] Sunil Sanka, Chittaranjan Hota, and Muttukrishnan Rajarajan. Secure

data access in cloud computing. In Internet Multimedia Services Architec-
ture and Application (IMSAA), 2010 IEEE 4th International Conference
on, pages 1–6, 2010. 25, 34, 77
[134] Laura Savu. Cloud computing: Deployment models, delivery models, risks
and research challenges. In Computer and Management (CAMAN), 2011
International Conference on, pages 1–4, 2011. 25, 36, 38, 39, 42, 52
[135] Hutton A. Meisel A. Windel A. Mohammed A. Licciardi A. Seccombe,
A. Security guidance for critical areas of focus in cloud computing, v2.1.
CloudSecurityAlliance, page 25, 2009. cited By (since 1996) 1. 9, 63
[136] N. K. Sehgal, Sohum Sohoni, Ying Xiong, David Fritz, Wira Mulia, and
J. M. Acken. A cross section of the issues and research activities related to
both information security and cloud computing. IETE Technical Review,
28(4):279, 2011. 11, 30, 33, 35, 38, 42, 44, 46, 57, 72
[137] F.B. Shaikh and S. Haider. Security threats in cloud computing. In In-
ternet Technology and Secured Transactions (ICITST), 2011 International
Conference for, pages 214 –219, December 2011. 13, 15, 18, 79
[138] Swarnpreet Singh and Tarun Jangwal. Cost breakdown of public cloud
computing and private cloud computing and security issues. International
Journal of Computer Science, 4. 33, 43, 47, 49, 52
[139] Avvari Sirisha and G. Geetha Kumari. API access control in cloud using
the role based access control model. In Trendz in Information Sciences &
Computing (TISC), 2010, pages 135–137, 2010. 37, 73
[140] Dawn Song, Elaine Shi, Ian Fischer, and Umesh Shankar. Cloud data
protection for the masses. Computer, 45(1):39–45, 2012. 58, 74
[141] D.X. Song, D. Wagner, and X. Tian. Timing analysis of keystrokes and
timing attacks on SSH. In Proceedings of the 10th conference on USENIX
Security Symposium-Volume 10, pages 25–25, 2001. 44
[142] Prashant Srivastava, Satyam Singh, Ashwin Alfred Pinto, Shvetank Verma,
Vijay K. Chaurasiya, and Rahul Gupta. An architecture based on proactive
model for security in cloud computing. In Recent Trends in Information
Technology (ICRTIT), 2011 International Conference on, pages 661–666,
2011. 27, 29, 32, 33, 34, 37, 38, 40, 45, 49, 57, 58, 59, 60, 63, 73
[143] Salvatore J. Stolfo, Malek Ben Salem, and Angelos D. Keromytis. Fog
computing: Mitigating insider data theft attacks in the cloud. In Security
and Privacy Workshops (SPW), 2012 IEEE Symposium on, pages 125–128,
2012. 31, 50, 75
References 120
[144] S. Subashini and V. Kavitha. A survey on security issues in service delivery

models of cloud computing. Journal of Network and Computer Applications,
34(1):1–11, January 2011. 13, 25, 26, 27, 28, 29, 30, 31, 36, 38, 41, 47, 48,
49, 50, 53, 59, 65, 66, 67, 68, 69
[145] H. Takabi and J.B.D. Joshi. StateMiner: an efficient similarity-based ap-

proach for optimal mining of role hierarchy. In Proceeding of the 15th ACM
symposium on Access control models and technologies, pages 55–64, 2010.
68, 70
[146] H. Takabi, J.B.D. Joshi, and G. Ahn. Security and privacy challenges
in cloud computing environments. Security & Privacy, IEEE, 8(6):24–31,
2010. 8, 11, 45, 46, 50, 59, 64, 65, 66, 68, 70, 71
[147] H. Takabi, M. Kim, J.B.D. Joshi, and M.B. Spring. An architecture for
specification and enforcement of temporal access control constraints using
OWL. In Proceedings of the 2009 ACM workshop on Secure web services,
pages 21–28, 2009. 70
[148] Hassan Takabi, James BD Joshi, and Gail-Joon Ahn. Securecloud: Towards
a comprehensive security framework for cloud computing environments.
In Computer Software and Applications Conference Workshops (COMP-
SACW), 2010 IEEE 34th Annual, pages 393–398, 2010. 46, 48, 49, 55,
77
[149] Yuan Tian, Biao Song, Jimupimg Park, and Eui-Nam Huh. Inter-cloud data
integration system considering privacy and cost. Computational Collective
Intelligence. Technologies and Applications, pages 195–204, 2010. 57, 75
[150] L.M. Vaquero, L. Rodero-Merino, J. Caceres, and M. Lindner. A break

in the clouds: towards a cloud definition. ACM SIGCOMM Computer
Communication Review, 39(1):50–55, 2008. 3
[151] Luis M. Vaquero, Luis Rodero-Merino, and Daniel Morán. Locking the sky:
a survey on IaaS cloud security. Computing, 91:93–118, November 2010. 56
[152] J. Varia. Building GrepTheWeb in the cloud, part 1: Cloud architectures.

Forschungsbericht, Amazon Web Services, URL http://developer. amazon-
webservices. com/connect/entry. jspa. 67
[153] M. Venkatesh, M. R. Sumalatha, and C. SelvaKumar. Improving public

auditability, data possession in data storage security for cloud computing.
In Recent Trends In Information Technology (ICRTIT), 2012 International
Conference on, pages 463–467, 2012. 28, 34, 74, 75
References 121
[154] J. Viega. Cloud computing and the common man. Computer, 42(8):106–8,
2009. Copyright 2009, The Institution of Engineering and Technology. 25,
26, 47
[155] C. Wang, Q. Wang, K. Ren, and W. Lou. Ensuring data storage security in
cloud computing. In Quality of Service, 2009. IWQoS. 17th International
Workshop on, pages 1–9, 2009. 65
[156] Cong Wang, Ning Cao, Jin Li, Kui Ren, and Wenjing Lou. Secure ranked
keyword search over encrypted cloud data. In Distributed Computing Sys-
tems (ICDCS), 2010 IEEE 30th International Conference on, pages 253–
262, 2010. 27, 77
[157] L. Wang, G. Von Laszewski, A. Younge, X. He, M. Kunze, J. Tao, and

C. Fu. Cloud computing: a perspective study. New Generation Computing,
28(2):137–146, 2010. 8, 9
[158] Suzanne C. Watson. A primer in survey research. The Journal of Continuing

Higher Education, 46(1):31–40, 1998. 86, 88
[159] S. William. Network Security Essentials. Pearson Education India, 2008.

38, 39
[160] Piers Wilson. Positive perspectives on cloud security. Information Security

Technical Report, 2011. 25, 31, 45, 57
[161] Claes Wohlin, Martin Höst, and Kennet Henningsson. Empirical research
methods in software engineering. In Empirical Methods and Studies in
Software Engineering, volume 2765 of Lecture Notes in Computer Science,
pages 7–23. Springer Berlin / Heidelberg, 2003. 20
[162] K. Wood and E. Pereira. An investigation into cloud configuration and

security. In Internet Technology and Secured Transactions (ICITST), 2010
International Conference for, pages 1–6. IEEE, November 2010. 32, 33, 34,
50, 52
[163] Ruoyu Wu, Gail-Joon Ahn, Hongxin Hu, and Mukesh Singhal. Information
flow control in cloud computing. In Collaborative Computing: Networking,
Applications and Worksharing (CollaborateCom), 2010 6th International
Conference on, pages 1–7, 2010. 3, 43, 50, 75
[164] Zhang Xin, Lai Song-qing, and Liu Nai-wen. Research on cloud computing
data security model based on multi-dimension. In Information Technology
in Medicine and Education (ITME), 2012 International Symposium on,
volume 2, pages 897–900, 2012. 25, 32, 34, 37, 38, 46, 52, 60, 74
References 122
[165] Liang Yan, Chunming Rong, and Gansen Zhao. Strengthen cloud comput-
ing security with federal identity management using hierarchical Identity-
Based cryptography. In Proceedings of the 1st International Conference on
Cloud Computing, CloudCom ’09, pages 167–177, Berlin, Heidelberg, 2009.
Springer-Verlag. 24, 49
[166] Zhang Yandong and Zhang Yongsheng. Cloud computing and cloud security
challenges. In Information Technology in Medicine and Education (ITME),
2012 International Symposium on, volume 2, pages 1084–1088, 2012. 29,
31, 34, 37, 41, 43, 49, 52, 59
[167] Jianfeng Yang and Zhibin Chen. Cloud computing research and security
issues. In Computational Intelligence and Software Engineering (CiSE),
2010 International Conference on, pages 1–3, 2010. 8
[168] L. Youseff, M. Butrico, and D. Da Silva. Toward a unified ontology of cloud

computing. In Grid Computing Environments Workshop, 2008. GCE’08,
pages 1–10, 2008. 35
[169] Huiming Yu, Nakia Powell, Dexter Stembridge, and Xiaohong Yuan. Cloud
computing and security challenges. In 50th Annual Association for Comput-
ing Machinery Southeast Conference, ACM-SE’12, March 29, 2012 - March
31, 2012, Proceedings of the Annual Southeast Conference, pages 298–302.
Association for Computing Machinery, 2012. 25, 27, 36, 37, 38, 39, 40, 52,
73
[170] Shucheng Yu, Cong Wang, Kui Ren, and Wenjing Lou. Achieving secure,
scalable, and fine-grained data access control in cloud computing. In IN-
FOCOM, 2010 Proceedings IEEE, pages 1–9, 2010. 3, 25, 30, 48, 77, 78
[171] Zhiwei Yu, Chaokun Wang, Clark Thomborson, Jianmin Wang, Shiguo
Lian, and Athanasios V. Vasilakos. A novel watermarking method for soft-
ware protection in the cloud. Software: Practice and Experience, 42(4):409–
430, 2011. 51, 78
[172] M. Zeller, R. Grossman, C. Lingenfelder, M.R. Berthold, E. Marcade,

R. Pechter, M. Hoskins, W. Thompson, and R. Holada. Open standards
and cloud computing: Kdd-2009 panel report. In Proceedings of the 15th
ACM SIGKDD international conference on Knowledge discovery and data
mining, pages 11–18, 2009. 56
[173] L.J. Zhang and Q. Zhou. CCOA: cloud computing open architecture. In
Web Services, 2009. ICWS 2009. IEEE International Conference on, pages
607–616, 2009. 54
References 123
[174] Qi Zhang, Lu Cheng, and Raouf Boutaba. Cloud computing: state-of-the-

art and research challenges. Journal of Internet Services and Applications,
1(1):7–18, April 2010. 11
[175] Y. Zhang and J.B.D. Joshi. Access control and trust management for emerg-
ing multidomain environments. Information Assurance, Security and Pri-
vacy Services, pages 421–455, 2009. 68
[176] Han Zheng, Zhongmeng Zhao, and Xuanping Zhang. Access control based
on group signatures in cloud service. In Computer Science and Automation
Engineering (CSAE), 2012 IEEE International Conference on, volume 2,
pages 316–320, 2012. 73
[177] Miao Zhou, Yi Mu, Willy Susilo, Jun Yan, and Liju Dong. Privacy en-
hanced data outsourcing in the cloud. Journal of Network and Computer
Applications, 2012. 57, 58, 76
[178] W. Zhou, W.R. Marczak, T. Tao, Z. Zhang, M. Sherr, B.T. Loo, and I. Lee.
Towards secure cloud data management. Technical Reports (CIS), page 919,
2010. 25, 32, 63, 78
[179] Dimitrios Zissis and Dimitrios Lekkas. Addressing cloud computing security
issues. Future Generation Computer Systems, (0), 2010. 28, 46, 53, 57, 69
Appendix A
Initial Set of papers
Table A.1: Initial set of papers
Sl.no Papers Authors

responses
(yes/no)
1 A survey on security issues in service delivery mod- No
els of cloud computing
2 Addressing cloud computing security issues Yes
3 An Investigation into Cloud Configuration and Se- Yes
curity
4 Can a Trusted Environment Provide Security? No
5 Cloud Computing Implementation, Management, No
and Security
6 Cloud Computing: Issues and Challenges No
7 Cloud computing security issues and challenges Yes
8 Infrastructure as a Service Security: Challenges Yes
and Solutions
9 Locking the sky: a survey on IaaS cloud security No
10 Privacy and Anonymization as a Service: PASS No
11 Security and Privacy Challenges in Cloud Com- Yes
puting Environments
12 Security Challenges in Cloud Computing No
13 Storm Clouds Rising: Security Challenges for IaaS No
Cloud Computing
124
Appendix B
Survey questionnaire
The survey questions are distributed through an online survey link. This link
is prepared with 2 pages. The first page is to gather basic information and the
second page to gather solution/practices/guidelines.
Page One:
General information (OPTIONAL):
Name:
Designation:
Organization:
Total work experience in cloud:
Email-id:
Question 1: Who is your service provider for cloud computing applications?

a) Microsoft
b) Google
c) Amazon
d) Salesforce.com
e) Rackspace
Other, please specify:
Question 2: Types of services used

a) SaaS
b) PaaS
c) IaaS
Question 3: Types of cloud used
125
Appendix B. Survey questionnaire 126
a) Public cloud
b) Private cloud
c) Hybrid cloud
d) Community cloud
Page 2
Thank you for sharing your basic information.

We are Master students in Software Engineering from Blekinge Institute of
Technology (BTH), Sweden. Currently working on our Master thesis in Cloud
Computing security under the guidance of Prof. Lars Lundberg, BTH, Sweden.
This survey aims to collect expert opinion and experiences on how to mitigate
the issues identified under ‘incompatibility’. We have 3 major concerns identified
under this challenge and they are:
1. Interoperation between different service providers,
2. Migration of data, application from one service provider to another,
3. Compatibility of traditional (on-premise) authentication methods with cloud
security systems.
Following are few questions related to these issues and we request you share
your opinion and experiences to mitigate them.
Please note that these questions in no way require any business information
from you. We just need your OPINION and EXPERT advice.
1. Using resources from different cloud service providers collectively could help
a company achieve improved scalability and performance. For example an
application designed on Microsofts Azure platform using Oracle or Amazon
storage services or vice versa. This interoperation between different service
providers is a challenge in cloud computing.
a) What steps can be followed to provide interoperability between two or

more service providers?
Character Count 1000

b) What precautions can be taken to protect client data while interfacing
between multiple cloud service providers?

Appendix B. Survey questionnaire 127
2. Company might want to migrate its data/applications from one service

provider to another. This migration from one cloud to another is also
a challenge, as while transferring the new service provider must not only
transfer data but also the security policies enforced on client data. If mi-
gration is not allowed data can be locked-in with the service provider and
might lead to conflicts.
(a) If service provider allows migration, what steps (methods used, pre-
cautions taken or guidelines) can be followed to provide easy migration
without compromising security?

(b) If service provider does not allow migration, what precautions need to
be followed to avoid data lock-in?
3. For any user to access the company data/applications he/she must be au-
thenticated by an Identity management (IDM) system. Commonly used
authentication ways are Microsoft Windows authentication, Single Sign-On
(SSO), Light weight Directory Access Protocol (LDAP), Security Assertion
Mark up Language (SAML), OPENID and OAUTH, OpenSocial, Facebook
Connect, etc. Some of these methods are not compatible with cloud comput-
ing systems. If the enterprise IDM is not compatible with cloud platforms,
providing access to employees for different cloud applications can become
challenge. Hence delaying in the adoption of cloud.
a) What steps can be employed to provide compatibility between on-

premise authentication systems and cloud computing systems?

b) What on-premise authentication system compatible with cloud com-
puting system do you use?

Appendix C
Respondents
Table C.1: Respondents notation
Notation Position Years of experience in

cloud computing
R1 Asst. Manager 0
R2 Director IT/SW Services 1
R3 Associate Consultant 18 Months In Atos and 7+ yrs
In IT
R4 Senior Software Engineer 2
R5 Jr. Programmer Datamatics 2 years
R6 BPM HEAD 2 YEARS
R7 Survey Programmer 2.5
R8 Associate Director 3 years
R9 CEO 3 years
R10 Founder and CEO 5
R11 Senior Technology Practice Manager 5 years
R12 Solution Architect 7 years
R13 Strategy & Governance Leader 8 years
R14 Chief Arch part of 6 of largest imple-
mentations in the industry.
Across platforms, engineering
and hosting.
R15 Sr.OTM Consultant very less
128

Cloud Comp PDF

Uploaded by

Copyright:

Available Formats

Cloud Comp PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cloud Comp PDF

Uploaded by

Copyright:

Available Formats

Master Thesis

A Study on Cloud Computing

Kalyan Sudia 870418-0556

Context: Scientific computing in the 21st century has evolved from

Keywords: Cloud computing, Security, Challenges, Solutions, In-

List of figures vii

5 Surveys: Design and results 86

7 Conclusions and Future Work 105

Appendix A Initial Set of papers 124

Appendix B Survey questionnaire 125

Appendix C Respondents 128

3.1 Research questions and Research methodologies . . . . . . . . . . 21

4.1 Business breach report blog . . . . . . . . . . . . . . . . . . . . . 31

5.1 Compatible on-premise authentication systems used with cloud . . 99

A.1 Initial set of papers . . . . . . . . . . . . . . . . . . . . . . . . . . 124

C.1 Respondents notation . . . . . . . . . . . . . . . . . . . . . . . . . 128

4.1 Process of selecting papers for study . . . . . . . . . . . . . . . . 23

5.1 Results of survey: Service providers used . . . . . . . . . . . . . . 89

Acronyms used in this document

Accronym Acronym expanded

Accronym Acronym expanded

Cloud Computing (CC) is an emerging technology that has abstruse connection

1.1 Aims and Objectives

• Identify the challenges that have no mitigation strategies defined.

• Collect solutions/guidelines/practices from organizations, for a challenge

• List out solutions/practices/guidelines to the cloud computing security chal-

1.2 Research questions

2. What solutions are present to handle security related cloud challenges?

3. How does an organization handle the identified security challenge, which

4. What are the solutions or guidelines to enhance security in CC where no

1.3 Methods used

1.4 Structure of report

Chapter 3 Research Design, explains an outline on methodologies used to infer

Chapter 4 Literature Review, explains detailed process followed in selecting,

Chapter 6 Discussion, explains a discussion on findings in this report (summa-

2.1 History of cloud computing

2.2 Defining the term cloud computing

2.3 Benefits of cloud computing

• Reduced Cost: Since cloud technology is implemented incrementally (step-

• Increased Storage: When compared to private computer systems, huge amounts

• Flexibility: Compared to traditional computing methods, cloud computing

• Greater mobility: Accessing information, whenever and wherever needed

• Shift of IT focus: Organizations can focus on innovation (i.e., implement-

2.4 Cloud Computing: Service models

• Software as a Service (SaaS): The service provider in this context pro-

manage or control the underlying cloud infrastructure (i.e. network, oper-

• Platform as a Service (PaaS): The service provider in this context pro-

• Infrastructure as a Service (IaaS): The consumer is provided with

• Privacy and Anonymization as a Service (PAAS): This service is

• Hardware as a Service (HaaS): The idea of buying a hardware or an

• Identity as a Service (IDaaS): This service is targeted for third party

• Security as a Service (SaaS): This service allows users to create their

2.5 Cloud computing: Deployment models

2.6 Importance of security in cloud computing

2.7 Related work in cloud computing security

• Popovi and Hocenski, discussed security issues, requirements and challenges

1. Security issues describe the problems encountered during implementa-

• Maggi and Zanero, addressed countermeasures (anti-viruses, intrusion de-

a. Users authentication: User authentication process must be improvised