International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169

Volume: 5 Issue: 5 54 58
______________________________________________________________________________________________

Online Signature Verification and Authentication using Smart Phones

Harshil Shah Pranav Pawar Mr. S.P. Khachane

Student, Department of Computer Student, Department of Computer Asst Professor, Dept of Computer
Engineering, MCT's Rajiv Gandhi Engineering, MCT's Rajiv Gandhi Engineering, MCT's Rajiv Gandhi
Institute Of Technology
Institute Of Technology Institute Of Technology
Mumbai, Maharashtra
Mumbai, Maharashtra pranav.pawar661@gmail.com Mumbai, Maharashtra
harshilshah1910@me.com khachnesp@gmail.com

Shikhar Sharma Shrey Pithava

Student, Department of Computer Engineering Student, Department of Computer Engineering
MCT's Rajiv Gandhi Institute Of Technology MCT's Rajiv Gandhi Institute Of Technology
Mumbai, Maharashtra Mumbai, Maharashtra
shikhar168@gmail.com shreygalaxy3@gmail.com

Abstract The proposed system is designed to determine whether the person signing on any touch screen device is authenticated user or not.
This can be done by verifying his/her handwritten signature which is a socially accepted biometric trait for authenticating an individual. There
are two types of handwritten signature verification systems: offline and online systems. In an off-line system, just an image of the users
signature is acquired without additional attributes, whereas, in an online system, a sequence of x-y coordinates of the users signature, along
with many other attributes are also acquired. In our paper, we have created a client (mobile) application which captures the users signature and
extracts various features like pressure, time and x-y co-ordinates and the server application verifies these features to find whether the signature
has been done by an authenticated user or a forger. The implementation is done using Python and the GUI is coded using Xcode.

Keywords- Handwritten signature, biometric, verification, authentication, offline, online, forger, Python, GUI, Xcode

__________________________________________________*****_________________________________________________

I. Introduction Identification can be done using a person's identity based only

Day by day, natural and secure access to interconnected on biometric measurements. The comparator matches the
systems is becoming more and more important. It is also obtained biometric with the ones enrolled in the database
necessary verifying people's identity in a fast, easy to use and using a 1: N matching algorithm for identification.Verification
user-friendly way.To achieve more reliable verification or involves the process of confirming or denying a person's
identification we should use something that uniquely claimed identity. The biometric data obtained from the user is
recognizes the given person. compared to the users data already stored in the database.
Authentication is the process of proving or verifying ones In comparison with other already adopted biometric
identity. It can be categorized in three types : something we verification techniques, signature verification presents many
know, like passwords; something we have, like bus tickets or likely advantages like nowadays it is a socially accepted
tokens; and, something we are, like our face, voice, signatures, method already in use in banks and credit card transaction.
etc. The third type is also known as Biometric. Together, these Also, it is useful for most of the new generation of portable
types are known as 3 factors of authentication[2]. computers and personal digital assistants (PDAs) use and
Biometrics means the automatic identification of a person writing as the main input channel. A signature may be
based on his/her physiological or behavioural changed by the user. Similarly to a password while it is not
characteristics[1]. This method of verification is preferred possible to change finger prints iris or retina patterns.
over traditional methods involving passwords and PIN
numbers for its accuracy and case sensitiveness. A biometric In signature verification, handwritten signature is commonly
system is essentially a pattern recognition system which used and accepted as a way to verify peoples
makes a personal identification by determining the identity.Signature verification usually consists just of an eye
authenticity of a specific physiological or behavioural inspection as if we compared two photographs, but this is not
characteristic possessed by the user. These characteristics are an efficient method against imposters. Document examiners
measurable and unique. commonly compare a suspect signature with several example
54
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 54 58
______________________________________________________________________________________________
of known valid signatures. They look for signs of forgery be regarded as one-factor authentication techniques, as they
which include i)signature written at higher speeds than only rely on one factor i.e. something you know. These
normal, ii)rounded endings and beginnings, iii)poor line authentications systems also might have Security Questions
quality with shaking of line and iv)stops in places where for the purpose of password recovery in case the user is unable
writing should be free[5]. to remember it[7]. These systems have many drawbacks.
All signatures that can be used for verification are stored in a Firstly, the user is required to remember a password, often a
signature database. The system uses a database, in which each separate password for each different service or website they
individual has 25 true signatures. At the same time, each use so that a password leak on one site doesnt leak all their
individual makes 5 forgeries of every of his/her 5 immediately passwords[6].
previous entries in the database. This means that for every The security of these passwords from brute-force attacks is
individual we have 25 true signatures and 25 forgeries made often dependent on the length of the password, which means
by 5 different people. Also the efficiency of this system is that the user often is forced to make a trade-off between
more as the forger tries several times to imitate the true users security (more complex and thus secure password) and
signature before the forgery is acquired and finally stored in convenience (simple and thus easy to remember, but less
the database. secure).
II. Literature Review
IV. Proposed system
In our proposed signature verification system, we take 25
Design Of Digital Signature Verification Algorithm Using
sample signatures as input from user in different conditions
Relative Slope Method: [3]
and then 25 forgeries. These are all stored in a database.
In this paper, signature is taken in 25 different conditions and
Normalisation is applied to signature made while logging in.
that is used to match with signature made while
HMM is used to segment the normalised sign and assign
authenticating. Normalisation, Hidden Markov Model and
probabilities. Relative slope algorithm is applied to calculate
Relative Slope Algorithm are used. Comparison of database
the relative slopes of input and stored sign and then they are
signature and input signature is made for authentication.
compared.
Because 25 different conditions are used for the samples, so
The algorithm uses simple geometric features to characterise
user can use the system in multiple conditions. Extra storage
signatures that effectively serve to distinguish signatures of
space is required to store 25 signatures.
different persons. We have a client (mobile) application which
Online Signature Verification For Multi-modal Authentication captures the users data and the server application verifies the
Using Smart Phones: [5] data. The implementation is done using Python and the GUI is
Here, we collect data(Sample signature) and will represent it coded using Xcode.
on x and y axis in every possible way which are upside down, V. Workflow of the system
take mirror image, take mirror image of upside down etc.
After representing signal, we will calculate frequency of x and
y coordinated in a 10x10 grid and also will calculate angle
made by line using any two points of sampled signal with x
axis. Algorithm used are Frequency String Method, Angle
String Method. Therefore, this approach shows effective and
easy way for feature extraction, easy to implement algorithm.
But this system requires more processing time.

Online signature Verification Using Mobile Devices: [4]

The technique used in this paper tries to collect more precise
data information such as considering time, speed, angle,
FIGURE 1 :- PHASE 1
pressure of signature Algorithms used are 1-Dimenstional
histograms, 2-Dimentional histograms. The advantage of this
paper is that more accurate and complex algorithms are
explained here which will make almost next to impossible for
an unauthorized person to login to the system.
III. Existing System
Most current login authentication systems such as one used in
websites tend to require a combination of either email or
username along with the users chosen password[1]. These can FIGURE 2 :- PHASE 2
55
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 54 58
______________________________________________________________________________________________
7) The second step used to calculate the length of signature
completed in unit time for this two tier time metric extraction
algorithm is used.
8) And finally in verification step two level verification
algorithm used in first level of verification relative slope value
compare with previous value
9) And the second level extracts the global and local features
FIGURE 3:-PHASE 3
consisting of relative slope value, total length and global time.
The above figures show the workflow of the proposed system
10) If the signature passes the second level of verification it
in phases.
considered as genuine signature
Phase 1:
Hidden Markov Model(HMM) :-
In user registration process user credentials is recorded, and
A hidden Markov model (HMM)[1] is a statistical Markov
along with the signature performed by the user in multiple
model in which the system being modelled is assumed to be a
conditions it is stored in the database. For signature storage,
Markov process with unobserved (hidden) states. In a hidden
first the user is made to perform signature in multiple different
Markov model, the state is not directly visible, but the output,
conditions. Then, normalization is performed on the
dependent on the state, is visible. Each state has a probability
signatures, and this data is then stored in the database.
distribution over the possible output tokens[8]. Therefore, the
sequence of tokens generated by an HMM gives some
Phase 2:
information about the sequence of states.
It consists of forgery prevention process in which we show
users signature to 5 professional forgers, each of which
performs 5 forged signatures of the users real signature. This
data is then stored in the database.

Phase 3:
This phase involved verification of signature in the login
process. In this, user enters credentials and signature to verify.
Figure 4:- Markov Chain
This signature is then compared with the real and forged
signatures as stored in the database. If the signature is verified
A Markov chain given in above figure 4 consist of two states,
as authentic, the user is logged in.
one is known as hidden states and other is known as
The proposed system is majorly implemented by using 2 observable states. Every observable state is connected to its
respective hidden state which we need to find. Looking at the
algorithms:- Relative Slope Algorithm[3] and Hidden Markov
observable state we predict the hidden state, this is the essence
Model[1][2].
of both Hidden Markov model and Markov chain.
Relative Slope Algorithm(RSA):- Also there are 2 probabilities in HMM as given below:-
This algorithm calculates all the slope values of the stored 1)Transition probability :-
the probability of going from a given state to the next state in a
signature[3]. The steps to be followed are given below:-
Markov process.(probability of hidden state)
Various steps to be followed are given below:
2)Observe probability
1) Using optimize HMM we can calculate the segment of the
signature. Then segment can be combining to form a line It is the probability of any visible state.
segment.
The basic idea of implementation of HMM in our paper is
2) After the line segment are obtained the relative slope are
assigning initial probability to each state. During the process
calculated.
of feature extraction we will extract various features like
3) Slope of line: S=dy/dx Where: dx=x2-x1 dy= y2-y1
4) For the first segment we calculate the slope between the pressure, time, slope and x-y co-ordinates. As the slope aspect
starting point of the first segment and the ending point of the has been covered by above algorithm we are considering
pressure for implementation of markov model. Different
last segment.
readings of pressure form different state of markov model and
5) However, the for the further line segment the slope is
to each state we assign a initial probability and and also
calculated based on the previous line segment.
6) In the first step global time required to put the signature and calculate the transition probability to each state. This can be
calculate. done by following the steps given below:-

56
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 54 58
______________________________________________________________________________________________
1) The very first step is to run a loop through the available
dataset and find the max value and min value
2) After calculating min and max value use formula (max -
min)/5 to divide the hole dataset into 5 ranges
3) Then again run a loop determine the number of time a data
set is occurring in that range divided by the total number
of datasets.
4)Transitional probability=(Occurrence in particular range/
Total datasets)
5) Now initial probability is assigned according to the formula
(Occurrence in particular range/ 5)
6) Now this process is repeated for both forged and real
datasets of a particular user and value is stored
7) Now the real time signature of user is present state and all
calculate ranges are act as transitional states(Hidden States) or
all set of possible state in which the system can make
transition to. Figure 6:- Result when a forger signs
8) Using above states we predict whether the signature is of
real user or forged user(Observational state).

VI. Implemented System

The system uses a GUI to capture 25 signatures of
authenticated user as well as 25 signatures of forgers. While
implementing, the signatures were done by 1 authenticated
user and 3 forgers. Figure 5 shows the signature of the
authenticated user being accepted. Whereas figures 6, 7 and 8
shows the signatures of the forgers being rejected. Due to the
normalization algorithm[7] the user can sign anywhere on the
screen and the result would remain unaffected by it.

Figure 7:- Result when a forger signs

Figure 5:- Result when authenticated user signs

Figure 8:- Result when a forger signs

57
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 54 58
______________________________________________________________________________________________
VII. Conclusion & future work [2] Javier Ortega-Garcia, Daniel Ramos- "HMM-Based On-
In the current times of increasing forgeries, it is important to Line Signature Verification: Feature Extraction &
use a system which will help to efficiently manage user Signature Modeling"
[3] P.N .Ganorkar, Kalyani Pendke - Design Of Digital
authentication in a more systematic and secure manner.
Signature Verification Algorithm Using Relative Slope
Criminal experts can be employed at every place and hence
Method - eISSN: 2319-1163 | pISSN: 2321-7308, August
there is an increase in need to develop computerized 14
algorithms that could verify and authenticate individuals [4] Napa Sae-Bae, Nasir D Menon- Online Signature
identity. Our mobile based authentication system can be Verification on mobile devices - ISSN 2250-2459, June
considered as a complete solution for multi-factor 2014
authentication. This is because we extract features of the users [5] Navid Forhad, Bruce Poon, M. Ashraful Amin, Hong Yan -
in real time and have created a training model with the help of Online Verification for multi-modal Authentication using
3-4 forgers which will cover various forgeries that could be Smart Phone -ISSN: 2078-0958, Volume 1, March 2015
[6] Kiran Kumer Gurrala, Sukadev Meher- Online Signature
done by many professional forgers.
Verification Techniques - ISSN: 2319-7064, November 13
In future, we would like to work on increasing the systems
[7] Aswathy K. V. - Online Signature Verification
accuracy so that it can perform at its best even in 1-sigma Techniques: A Survey, ISSN 2091-2730
range[8]. To increase the accuracy we could implement many [8] Beton, M., Marie, V., Rosenberger, C.: Biometric secret
more algorithm such as optimized HMM, Angle deviation path for mobile user authentication: A preliminary study.
method[7] also we can reduce the size of the database as 25 In: 2013 World Congress Computer Information
copies of signature each for real and forger will occupy lot of Technology, pp. 16 (2013)
database.
VIII. References
[1] Juan J. Igarza, Iaki Goirizelaia - "Online Handwritten
Signature Verification Using Hidden Markov Model"

58
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________