Forensic Analysis of Whatsapp Messenger On Android Smartphones

Forensic Analysis of WhatsApp Messenger
on Android Smartphones
arXiv:1507.07739v1 [cs.CR] 28 Jul 2015
Cosimo Anglano
DiSIT - Computer Science Institute,
Universitá del Piemonte Orientale, Alessandria (Italy)
email:cosimo.anglano@uniupo.it
This manuscript is the preprint of the paper

Forensic Analysis of WhatsApp Messenger on Android
Smartphones,
that has been published on the Digital Investigation Journal,
Vol. 11, No. 3, pp. 201–213, September 2014
doi:10.1016/j.diin.2014.04.003
(c) 2014. This manuscript version is made available under the
CC-BY-NC-ND 4.0 license
http://creativecommons.org/licenses/by-nc-nd/4.0/
Abstract
We present the forensic analysis of the artifacts left on Android
devices by WhatsApp Messenger, the client of the WhatsApp instant
messaging system. We provide a complete description of all the arti-
facts generated by WhatsApp Messenger, we discuss the decoding and
the interpretation of each one of them, and we show how they can be
correlated together to infer various types of information that cannot
be obtained by considering each one of them in isolation.
By using the results discussed in this paper, an analyst will be
able to reconstruct the list of contacts and the chronology of the mes-
sages that have been exchanged by users. Furthermore, thanks to the
correlation of multiple artifacts, (s)he will be able to infer informa-
tion like when a specific contact has been added, to recover deleted
contacts and their time of deletion, to determine which messages have
1
been deleted, when these messages have been exchanged, and the users
that exchanged them.
1 Introduction
The introduction of sophisticated communication services over the Internet,
allowing users to exchange textual messages, as well as audio, video, and
image files, has changed the way people interact among them. The usage
of these services, broadly named instant messaging (IM), has undoubtedly
exploded in the past few years, mainly thanks to the pervasiveness of smart-
phones, that provide quite sophisticated IM applications. Smartphones in-
deed enable users to exploit their data connection to access IM services any-
where and anytime, thus eliminating the costs usually charged by mobile
operators for similar services (e.g., for SMS communication).
Given their popularity, IM services are being increasingly used not only
for legitimate activities, but also for illicit ones [20]: criminals may indeed use
them either to communicate with potential victims, or with other criminals
to escape interception [3]. Therefore, IM applications have the potential of
being a very rich source of evidentiary information in most investigations.
Among IM applications for smartphones, WhatsApp [24] is accredited to
be the most widespread one (reportedly [25], it has over 400 million active
users that exchange, on average, more than 31 billion messages per day, 325
millions of which are photos [12]). Given its recent acquisition by Face-
book, it is reasonable to expect a further growth in its diffusion. Therefore,
the analysis of WhatsApp Messenger, the client of WhatsApp that runs on
smartphones, has recently raised the interest of the digital forensics commu-
nity [19, 10, 21].
In this paper we deal with the forensic analysis of WhatsApp Messenger
on Android smartphones. Android users, indeed, arguably represent the
largest part of the user base of WhatsApp: as of Jan. 2014, Google Playstore
reports a number of downloads included between 100 and 500 millions (the
lower limit having been already hit in Nov. 2012), out of a population of 400
millions of users. Thus, by focusing on the Android platform, we maximize
the potential investigative impact of our work.
Several works, appeared recently in the literature [19, 10] deal with the
same problem. However, as discussed later, these works are limited in scope,
as they focus on only the reconstruction of the chronology of exchanged
2 available under the CC-BY-NC-ND

(c) 2014. This manuscript version is made
4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/
messages, and neglect other important artifacts that, instead, are considered
in our work.
More precisely, the contributions of this paper can be summarized as
follows:
• we discuss the decoding and the interpretation of all the artifacts and
data generated by WhatsApp Messenger on Android devices;
• we show how these artifacts can be correlated together to infer various
types of information that cannot be obtained by considering each one of
them in isolation, such as when a contact has been added to or deleted
from the contacts database, whether a message has been actually de-
livered to its destination after having been sent or has been deleted, if
a user joined or left a group chat before or after a given time, when a
given user has been added to the list of contacts, etc..
The rest of the paper is organized as follows. In Sec. 2 we review existing
work, while in Sec. 3 we describe the methodology and the tools we use in
our study. Then, in Sec. 4 we discuss the forensic analysis of WhatsApp
Messenger and, finally, in Sec. 5 we conclude the paper and outline future
research work.
2 Related works
The forensic analysis of IM applications on smartphones has been the subject
of various works published in the literature.
Compared with existing works, however, our contribution (a) has a wider
scope, as it considers all the artifacts generated by WhatsApp Messenger
(namely, the database of contacts, the log files, the avatar pictures, and the
preference files), (b) presents a more thorough and complete analysis of these
artifacts, and (c) explains how these artifacts can be correlated to deduce
various type of information having an evidentiary value, such as whether a
message has been actually delivered to its destination after having been sent,
if a user joined or left a group chat before or after a given time, and when a
given user has been added to the list of contacts.
[8] focus on the forensic analysis of three IM applications (namely AIM,
Yahoo! Messenger, and Google Talk) on the iOS platform. Their work
differs from ours for both the IM applications and the smartphone platform
it considers.

[2] focus on the analysis of several IM applications (including WhatsApp
Messenger) on various smartphone platforms, including Android, with the
aim of identifying the encryption algorithms used by them. Their work,
unlike ours, does not deal with the identification, analysis, and correlation of
all the artifacts generated by WhatsApp Messenger.
[21] focus on the analysis of iTunes backups for iOS devices with the
aim of identifying the artifacts left by various social network applications,
including WhatsApp Messenger. Their work differs from ours because of its
focus on iTunes and iOS, and because only the chat database of WhatsApp
is considered, since only this artifact is included into an iTune backup. Fur-
thermore, the information stored into the chat database is analyzed only in
part.
The works of [19] and [10] are similar to ours, since they both focus on the
forensic analysis of WhatsApp Messenger on Android. However, these works
focus mainly on the forensic acquisition of the artifacts left by WhatsApp
Messenger, and deal with their analysis only in part (they limit their study
to the chat database, and analyze it only partially). Similar considerations
apply to the WhatsApp Xtract tool [17], that extracts some of the informa-
tion stored into the chat database (and, possibly, in the contacts database),
without however providing any description of how these databases are parsed.
3 Analysis methodology and tools

The study described in this paper has been performed by carrying out a
set of controlled experiments, each one referring to a specific usage scenario
(one-to-one communication, group communication, multimedia message ex-
change, etc.), during which typical user interactions have taken place. After
each experiment, the memory of the sending and receiving devices has been
examined in order to identify, extract, and analyze the data generated by
WhatsApp Messenger in that experiment.
As discussed in Sec. 4, most of the files generated by WhatsApp Mes-
senger are stored into an area of the internal device memory that is nor-
mally inaccessible to users. To access this area, suitable commercial tools
may be used [4, 11, 14] but, unfortunately, we did not have access to them.
Open-source software-based tools are also available [7, 23], but we consider
problematic their use in our study for the following reasons:
• they may alter the contents of the memory, thus overwriting pieces of

information: while this can be considered acceptable in a real-world in-
vestigation when there is no other alternative, we believe that it should
be avoided in a study like the one presented in this paper, as modified
or incomplete data may yield to incorrect conclusions;
• they are device-specific: this would prevent a third party to replicate

the experiments to validate our findings, unless the same device model
and the same software acquisition tool are used.
For these reasons, in our study we adopted a different approach, in which

we use software-emulated Android devices in place of physical ones. In partic-
ular, we use the YouWave virtualization platform [26] that is able to faithfully
emulate the behavior of a complete Android device. YouWave implements
the internal device memory as a VirtualBox storage file [13], whose format is
documented and, therefore, can be parsed by a suitable tool to extract the
files stored inside it. In this way, the acquisition of the internal memory of
the device is greatly simplified, as it reduces to inspect the content of this
file.
In order to ensure the soundness of our approach, we have made tests
in which the behavior of, and the data generated by, WhatsApp Messenger
running on YouWave have been compared against those produced when it
runs on real smartphones. These tests have been performed either indirectly,
by comparing the data found in the inaccessible memory area of YouWave
against those documented in the literature[19, 10], or directly, by comparing
the data stored on the emulated SD memory card against those generated on
a real smartphone. The results of our tests indicate that, from the perspective
of WhatsApp Messenger, YouWave and a real smartphone behave the same
way.
Our experimental test-bed consists thus into a set of YouWave virtual
machines, namely one for each device involved in the experiments, running
Android v. 4.0.4. On each one of these machines we install and use What-
sApp Messenger v. 2.11. In each experiment, we assign a role to each virtual
device (e.g. sender or recipient of a message, group chat leader, etc.), and
use it to carry out the corresponding activities. Then, at the end of the
experiment, we suspend the virtual device, parse the file implementing the
corresponding internal memory (named youwave vm01.vdi) by means of FTK
Imager (v. 3.1) [1], and extract the files where WhatsApp Messenger stores

the data it generates. 1 These files are then examined by means of suit-
able tools. In particular, we use SqliteMan [22] to examine the databases
maintained by WhatsApp Messenger (as discussed later, they are SQLite v.3
databases [18]), and notepad++ [6] to examine textual files.
By proceeding as exposed above, (a) we are able to avoid the risks of
contamination and of an incomplete acquisition of the data stored in the
memory of the device, (b) we ensure repeatability of experiments, as their
outcomes do not depend on the availability of a specific software or hardware
memory acquisition tool or smartphone model, (c) we obtain a high degree
of controllability of experiments, as we may suspend and resume at will the
virtual device to perform acquisition while a given experiment is being carried
out and, last but not less important, (d) we reduce the costs of the study,
since neither real smartphones nor commercial memory acquisition tools are
necessary to carry out the experiments.
4 Forensic analysis of WhatsApp Messenger

WhatsApp provides its users with various forms of communications, namely
user-to-user communications, broadcast messages, and group chats. When
communicating, users may exchange plain text messages, as well as multime-
dia files (containing images, audio, and video), contact cards, and geolocation
information.
Each user is associated with its profile, a set of information that includes
his/her WhatsApp name, status line, and avatar (a graphic file, typically
a picture). The profile of each user is stored on a central system, from
which it is downloaded by other WhatsApp users that include that user in
their contacts. The central systems provides also other services, like user
registration, authentication, and message relay.
As reported in [19], the artifacts generated by WhatsApp Messenger on
an Android device are stored into a set of files, whose name, location, and
contents are listed in Table 1.
1
The only exception we make to the above methodology is the use of a physical smart-
phone to generate messages carrying geolocation coordinates, as the Android Location
Services, used by WhatsApp Messenger to obtain the coordinates of the current location
of the device, are not available on YouWave because of its lack of a GPS receiver. In this
case, access to the relevant data is achieved by using the backup mechanisms described in
Sec. 4.2.1.

Row Content Directory File
#
1 contacts /data/data/ wa.db (SQLite v.3)
database com.whatsapp/databases
2 chat database /data/data/ msgstore.db (SQLite v.3)
com.whatsapp/databases
3 backups of the /mnt/sdcard/ msgstore.db.crypt
chat database Whatsapp/Databases msgstore-<date>.crypt
4 avatars of /data/data/ UID.j, where UID is the
contacts com.whatsapp/files/ identifier of the contact
Avatars
5 copies of /mnt/sdcard/ What- UID.j, where UID is the
contacts avatars sApp/ProfilePictures identifier of the contact
6 log files /data/data/ whatsapp.log,
com.whatsapp/files/ whatsapp-<date>.log
Logs
7 received files /mnt/sdcard/ various files
Whatsapp/Media
8 sent files /mnt/sdcard/ various files
Whatsapp/Media/Sent
9 user settings /data/data/ various files
and preferences comm.whatsapp/files
Table 1: WhatsApp Messenger artifacts.

In the rest of this section we discuss how the above artifacts can be an-
alyzed and correlated to ascertain various types of information: we start
with contact information (Sec. 4.1), we continue with exchanged messages
(Sec. 4.2), and we end with application settings and user preferences (Sec. 4.3).
4.1 Analysis of contact information

The evidentiary value of contact information is notorious, as it allows an
investigator to determine who the user was in contact with.
In this section we first describe the information that are stored in the
contacts database, and then we discuss how this information can be analyzed
to determine (a) the list of contacts, (b) when a contact has been added to
the database, (c) whether and when a given contact has been blocked and,
finally, we show how deleted contacts can be dealt with.
4.1.1 Retrieving contact information

The contacts database wa.db contains three tables, namely wa contacts, that
stores a record for each contact, android metadata, and sqlite sequence, both
storing housekeeping information having no evidentiary value.
The structure of the records in wa contacts is shown in Table 2, where we
distinguish the fields containing data obtained from the WhatsApp system
(and, as such, having potential evidentiary value), from those storing data
extracted from the phonebook of the device (that, being set by the user and
not by WhatsApp, are not pertinent to our work).
As can be observed from this table, each record stores the WhatsApp ID
(field jid) of the contact, a string structured as ’x@s.whatsapp.net’, where ’x’ is
the phone number of that contact (for the sake of readability, in the following
we indicate users by means of their phone numbers instead of their complete
WhatsApp IDs). Furthermore, each record stores the profile name (field
wa name), and the status string (field status) of the corresponding contact.
Field is whatsapp user is instead used to differentiate actual WhatsApp users
from unreal ones: WhatsApp Messenger indeed adds to the contact database
a record for each phone number found in the phonebook of the device, even
if the corresponding user is not registered with the WhatsApp system.
Avatar pictures may have evidentiary value as well: they can be indeed
used to link a WhatsApp account to the real identity of the person using it
(for instance, if the avatar displays the face of the user, or any location or item

Data coming from the WhatsApp system
Field name Meaning
id sequence number of the record (set by SQLite)
jid WhatsApp ID of the contact (a string structured as
’x@s.whatsapp.net’, where ’x’ is the phone number of the
contact)
is whatsapp user contains ’1’ if the contact corresponds to an actual What-
sApp user, ’0’ otherwise
unseen msg count number of messages sent by this contact that have been
received, but still have to be read
photo ts unknown, always set to ’0’
thumb ts Unix epoch time (10 digits) indicating when the contact
has set his/her current avatar picture
photo id timestamp Unix millisecond epoch time (13 digits) indicating when
the current avatar picture of the contact has been down-
loaded locally
wa name WhatsApp name of the contact (as set in his/her profile)
status status line of the contact (as set in his/her profile)
sort name name of the contact used in sorting operations
Data coming from from the phonebook of the device
Field name Meaning
number phone number associated to the contact
raw contact id sequence number of the contact
display name display name of the contact
phone type type of the phone
phone label label associated to the phone number
given name given name of the user
family name family name of the user
Table 2: Structure of the wa contacts table

that can be uniquely associated with that person). The avatar picture of a
contact x@s.whatsapp.net is stored, as a JPEG file named x@s.whatsapp.net.j,
in the directories listed in Table 1, rows no. 4 and 5. The timestamps stored
in the thumb ts and photo id timestamp field indicate when the contacts has
set his/her current avatar, and when that avatar has been downloaded locally,
respectively.
4.1.2 Determining when a contact has been added

In some investigations, it may be necessary to determine when a given user
has been added to the contacts database 2 . This information is not stored in
the wa contacts table, but can be deduced from the analysis of the log files
generated by WhatsApp Messenger (that are located in the directory listed
in Table 1, row no. 6).
When a contact is added to the wa.db database, WhatsApp Messenger
logs several events that are tagged with their time of occurrence and with
the WhatsApp ID of the involved user.
Examples of these events, corresponding to the addition of user 39331xxxxxxx,
are reported in Fig. 1, from which we note that the following events are logged
each time a new user is added: (a) the discovery that the user is not present
yet in the contacts database (line no. 4), (b) the queries to the central system
to fetch various information about the contact (lines no. 7,10, and 14), and
(c) the completion of the download of the corresponding avatar picture (line
no. 17). From these events, we can determine when the user has been added
to the contacts database (on Sept. 25, 2013 at 14:14:24, in our example).
4.1.3 Dealing with blocked contacts

WhatsApp Messenger enables the user to block anyone of his/her contacts,
thus preventing any communication with him/her until the block is removed.
In an investigation it can be important to determine whether a contact was
blocked or not at a given time, in order to confirm or to exclude the reception
of a message sent at that time.
2
User contacts are automatically added to the contacts database by WhatsApp Mes-
senger that – each time is started or when the user starts a new conversation – inspects
the phonebook of the device and adds all the phone numbers that are not stored there
yet.

Figure 1: Events logged when a user is added to the contacts ta-
ble(phonenumber redacted to ensure the privacy of the owner). Long lines
have been split to ease readability.
The information concerning blocked users is stored neither in the contacts

database, nor elsewhere on the memory of the device (we conjecture that the
list of blocked contacts is stored on the WhatsApp central system, since when
the blocking is taking place, WhatsApp Messenger exchanges messages with
it). Blocked users can be however identified, under some circumstances, by
analyzing log files.
When a contact is blocked, an event – reporting the WhatsApp ID of
that contact and the time of occurrence of the operation – is indeed recorded
into the log file (see Fig. 2(a)). Unfortunately, when a contact is unblocked,
the event that is logged (Fig. 2(b)) does not report the WhatsApp IDs of the
involved contact, and it is cumulative (i.e., it may refer to a set of contacts
being unblocked simultaneously).
Thus, it is always possible to determine whether and when a given user
X has been blocked, but whether it is still blocked at a given time can be
ascertained only if either (a) no unblocking events are recorded in the log file
after the block operation, or (b) an unblocking event is present, but only user
X was blocked at that time. It follows that if several users are blocked and
one (or more) unblocking events are logged, it is not possible to tell which

Figure 2: Events in the log file corresponding to (a) the blocking, (b) the
unblocking of user 39320xxxxxxx.
users are still blocked, and which ones have been instead unblocked. It is
worth pointing out that the above inferences can be made only if the log files
reporting blocking and unblocking events are still available (i.e., they have
not been deleted by WhatsApp Messenger to create room for newer ones).
As a final consideration, we note that no information whatsoever is stored
on the side of the contact that gets blocked, so it is not possible to tell whether
the user of the device under analysis has been blocked or not by anyone of
his/her contacts.
4.1.4 Dealing with deleted contacts

In the attempt to hide past interactions, the user may delete a contact, thus
causing the removal of the corresponding record from the wa contacts table.
In some cases (notably, if the SQLite engine has not vacuumed the above
table yet), it may be possible to recover deleted records by means of suitable
techniques (e.g., [9, 16]. Our experiments, carried out by means of Oxygen
Forensic SQLite Viewer [15], indicate indeed that deleted contact records
may be recovered.
However, in general, at the moment of the analysis, deleted records may
have been vacuumed, so they cannot be recovered anymore. In these situ-
ations, it may be still possible to determine the set of deleted contacts by
first reconstructing the list of contacts that have been added in the past (by
analyzing log files as discussed in Sec. 4.1.2), and then by comparing this
list with the contents of the wa contacts table: the contacts in the list that
are not in the database are those that have been deleted. Note that this
procedure works only if the log file reporting the addition of a contact of
interest is still available when the analysis is performed.

Unfortunately, by proceeding as above, it is not possible to determine
when a given contact has been deleted, since deletions give rise to log events
that do not reference the WhatsApp ID of the contact being deleted.
4.2 Analysis of exchanged messages

WhatsApp Messenger stores all the messages that have been sent or received
into the chat database msgstore.db (located in the directory listed in Table 1,
row 2), whose analysis makes it possible to reconstruct the chronology of ex-
changed messages, namely to determine when a message has been exchanged,
the data it carried, the set of users involved in the conversation, and whether
and when it has actually been received by its recipients.
In the following we discuss each one of the above steps separately: we
start with the description of the structure of the chat database (Sec. 4.2.1),
and then we explain how to (a) reconstruct the chat history (Sec. 4.2.2), (b)
determine and extract the content of a message (Sec. 4.2.3), (c) determine
the status of a message (Sec. 4.2.4), (d) determine the set of users among
which each message has been exchanged (Sec. 4.2.5) and, finally, (e) deal
with deleted messages (Sec. 4.2.6).
4.2.1 The structure of the chat database

The msgstore.db database contains the following three tables:
• messages, that contains a record for each message that has been sent
or received by the user. To ease understanding, we classify the fields
of these records in two distinct categories: those storing attributes of
the message (listed in Table 3), and those storing the contents of the
message and the corresponding metadata (listed in Table 4);
• chat list, that contains a record for each conversation held by the user
(a conversation consists into the set of messages exchanged with a par-
ticular contact), whose fields are described in Table 5;
• sqlite sequence, that stores housekeeping data used internally by What-
sApp Messenger, whose structure is not reported here since it does not
have any evidentiary value.
As reported in [19], WhatsApp Messenger usually generates various backup
copies of the msgstore.db database, that are stored in the directory listed in

Field name Meaning
id record sequence number
key remote jid WhatsApp ID of the communication partner
key id unique message identifier
key from me message direction: ’0’=incoming, ’1’=outgoing
status message status: ’0’=received, ’4’=waiting on the
central server, ’5’=received by the destination,
’6’=control message
timestamp time of send if key from me=’1’, record insertion
time otherwise (taken from the local device clock,
and encoded as a 13-digits millisecond Unix epoch
time)
received timestamp time of receipt (taken from the local device clock,
and encoded as a 13-digits millisecond Unix epoch
time) if key from me=’0’, ’-1’ otherwise
receipt server timestamp time of receipt of the central server ack (taken from
the local device clock, and encoded as a 13-digits
millisecond Unix epoch time) if key from me=’1’, ’-
1’ otherwise
receipt device timestamp time of receipt of the recipient ack (taken from the
local device clock, and encoded as a 13-digits mil-
lisecond Unix epoch time) if key from me=’1’, ’-1’
otherwise
send timestamp unused (always set to ’-1’)
needs push ’2’ if broadcast message, ’0’ otherwise
recipient count number of recipients (broadcast message)
remote resource ID of the sender (only for group chat messages)
Table 3: Structure of the messages table: fields storing message attributes.

Field name Meaning
media wa type message type: ’0’=text, ’1’=image, ’2’=audio, ’3’=video,
’4’=contact card, ’5’=geo position)
data message content when media wa type = ’0’
raw data thumbnail of the transmitted file when me-
dia wa type={’1’,’3’}
media hash base64-encoded SHA-256 hash of the transmitted file (when
media wa type={’1’,’2’,’3’})
media url URL of the transmitted file (when me-
dia wa type={’1’,’2’,’3’})
media mime type MIME type of the transmitted file (when me-
dia wa type={’1’,’2’,’3’})
media size size of the transmitted file (when me-
dia wa type={’1’,’2’,’3’})
media name name of transmitted file (when media wa type={’1’,’2’,’3’})
media duration duration in sec. of the transmitted file (when me-
dia wa type={’1’,’2’,’3’})
latitude latitude of the message sender (when media wa type=’5’)
longitude longitude of the message sender (when media wa type=’5’)
thumb image housekeeping information (no evidentiary value)
Table 4: Structure of the messages table: fields storing information concern-

ing message contents.
Field name Meaning

id sequence number of the record
key remote jid WhatsApp ID of the communication partner
message table id sequence number of record in the messages table that corre-
sponds to the last message of the conversation
Table 5: Structure of the chat list table.

Table 1 row no. 3. These backups are full copies of the msgstore.db database,
and are not kept synchronized with it. Therefore, they are particularly im-
portant from an investigative standpoint, since they may store messages that
have been deleted from the main chat database. Backups are encrypted with
the AES 192 algorithm, but they can be easily decrypted since, as discussed
in [5], the same encryption key (namely, 346a23652a46392b4d73257c67317e352e3372482177652c)
is used on all devices.
4.2.2 Reconstruction of the chat history

To reconstruct the chronology of the messages exchanged by the user, the
records stored in the messages table must be extracted and decoded as dis-
cussed below.
To elucidate, let us consider Fig. 3, that shows four records correspond-
ing to a conversation between the device owner and the user 39348xxxxxxx
(actually, only the fields listed in Table 3 are displayed).
Figure 3: Reconstruction of the chat history. Phone numbers have been

grayed out to ensure the privacy of the owner.
By examining these records, we note that (a) all the messages have been
exchanged with the same contact 39348xxxxxxx (they all store the same What-
sApp ID in the key remote id field), (b) the conversation has been started
by that contact (key from me = ’0’ in record no.1) with a textual message
whose content was “Message 1” (field data) on Feb. 13th, 2012 06:59:09
(field received timestamp), and (c) the device owner replied at 07:00:23 of
the same day (field timestamp) with the message corresponding to record no.
2 (key from me=’1’) with content “Reply 1” (field data). The conversation
then continued with another message-reply exchange.
From these records, we also note that each message carries its own unique
identifier in the key id field: this value, set by the sender, is obtained by con-
catenating the timestamp corresponding to the last start time of WhatsApp

Messenger (on the sender’s device) with a progressive number (indicating the
number of messages sent from that moment), and is used also by the recipi-
ent to denote that message. Therefore, by using this value, it is possible to
correlate the records of the sender’s and recipient’s databases corresponding
to the same message.
4.2.3 Extracting the contents of a message

In addition to plain text messages, WhatsApp allows its users to exchange
messages containing data of various types, namely multimedia files (storing
images, audio, and video), contact cards, and geo-location information. The
type of data transmitted with a message is indicated (as reported in Table 4)
by the media wa type field, while the information concerning message content
is spread, for non-textual messages, over several fields (depending on the
specific data type). As a matter of fact, while the content of textual messages
(media wa type=’0’) is stored in the data field, for the other types of contents
the situation is more involved, as discussed below.
Multimedia files When the user sends a multimedia file, several activities
take place automatically (i.e., without informing the involved users). First,
WhatsApp Messenger copies the file into the folder listed in Table 1, row 8.
Then, it uploads the file to the WhatsApp server, that sends back the URL
of the corresponding location. Finally, the sender sends to the recipient a
message containing this URL and, upon receiving this message, the recipient
sends an acknowledgment back to the sender.
When these steps have been completed, the sender stores into his/her
messages table a record like the one shown in Fig. 4 (where we show only
the fields related to message contents that are listed in Table 4). As can
be seen from the above figure, the type of the file is indicated (besides the
wa media type field, not shown in the figure) by the media mime type field
(’image/jpeg’ in the example). Its name is instead stored in the media name
field (IMG-20131021-WA0000.jpg in the example), its size in bytes by me-
dia size (40267 in the example), and its thumbnail in the raw data field (as a
blob, i.e. a binary large object). Furthermore, the media url field stores the
URL of the location on the central server where the file has been temporar-
ily stored, whose last part (highlighted in Fig. 4 by framing it) corresponds
to the name given by the server to that file. Finally, the base64-encoded
SHA-256 hash of the transmitted file is stored in the media hash field.

Figure 4: Multimedia file exchange: sender side
On the recipient side, after message reception, the transmitted thumbnail

of the file is displayed by WhatsApp Messenger; the actual file is instead
downloaded at a later time only if the recipient explicitly requests it. Upon
message reception, the recipient stores in his/her messages table a record like
the one shown in Fig. 5. From this figure, we see that most fields are identical
Figure 5: Multimedia file exchange: recipient side
to those stored by the sender (in particular, wa media type, media mime type,
media size, raw data, and media hash). Conversely, the contents of media url
is different, except for the name given to the file by the server (highlighted
in Fig. 5 by framing it).
Unlike the sender, however, the media name field is empty, so the local
name given by WhatsApp Messenger to that file is unknown. The file can
be however identified by comparing the SHA-256 hash stored into the corre-
sponding record with that of all the files that have been received (that are

stored in the folder reported in Table 1, row 7.
Finally, we note that the file sent by the sender and the one received by
the recipient can be correlated by comparing both the file name given by the
WhatsApp server and the SHA-256 hash to these files (that are stored, as
discussed above, in the media url and media hash fields of the corresponding
records).
Contact cards Messages carrying contacts cards (extracted from the phone-
book of the sender) correspond – both on the sender and on the recipient
side – to messages record that store the transmitted information (in VCARD
format) into the data field, and the name given by the sender to that contact
in the media name field. An example of such a record is shown in Fig. 6.
Figure 6: Fields containing contact card information.
Geolocation coordinates WhatsApp Messenger enables users to send the

geographic coordinates of their current location, that are obtained from the
Android Location Services running on the device. Messages carrying geo-
graphic coordinates correspond –both on the sender and on the recipient
side – to messages records that store the latitude and the longitude values
into the latitude, longitude fields, and a JPEG thumbnail of the Google Map
displaying the above coordinates in the raw data field. An example of such a
record is shown in Fig. 7.
4.2.4 Determining the state of the message

In WhatsApp, messages are not exchanged directly among communicating
users, but they are first sent to the central server, that forwards them to the
respective recipients if they are on-line, and stores them locally until they
can be delivered otherwise. This implies that the presence of a record in the
messages table does not necessarily mean that an outgoing message has been

Figure 7: Geo-location message:(a) data stored in the database, (b) Google
map extracted from the raw data field.
actually delivered to its recipients. As a matter of fact, after the user has
pushed the “send” button of WhatsApp Messenger, the message can be in one
of the following three states: (a) waiting on the local device to be transmitted
to the central server, or (b) stored on the central server but waiting to be
transmitted to its recipient(s), or (c) delivered to its recipient(s).
The ability to distinguish the various states of a message may be crucial
in an investigation where it must be ascertained whether a message has been
actually delivered or not to its destinations, and when such a delivery has
taken place.
The current state of a message, as well as the times of its state changes,
can be understood by correlating the values contained in several fields of
the corresponding record of the sender database 3 , namely status, timestamp,
received timestamp, receipt server timestamp, and receipt device timestamp.
To explain, let us consider a scenario in which a user sends a message
when both him/her and the recipient are off-line (Fig. 8(a)), then the sender
gets reconnected to the network while the recipient is still offline (Fig. 8(b)),
and then, finally, also the recipient gets connected (Fig. 8(c)).
When the message is sent, a record is stored in the messages table of
the sender, even if the central server is unreachable. In this case, as shown
in Fig. 8(a), in this record we have that status=’0’, timestamp=’x’, and re-
ceived timestamp=’y’, where ’x’ and ’y’ correspond to when the user has
sent the message and when the record has been added to the chat database,
respectively.
Thus, a record such that key from me=’1’ and status=’0’ corresponds to
a message that has not been delivered to the central server yet.
3
For a recipient, a message can be only in the received state, corresponding to status=’0’

Figure 8: Sender side: record updates for a message while in transit.
Later, when the sender returns on-line, the message is forwarded to the
central server that replies with an ack. When this ack is received, the sender
updates the corresponding record as shown in Fig. 8(b) by setting status=’4’,
and the value of receipt server timestamp to the reception time of the ack.
a message that has been delivered the central server, but not yet to its desti-
nation(s).
Finally, when the recipient returns on line, it receives the message from
the central server, and sends an ack to the sender. Upon receiving this ack,
the sender updates again the record corresponding to that message (as shown
in Fig. 8(c)) by setting status=’5’, and the value of receipt device timestamp
to the reception time of the ack.
a message that has been delivered to its destination.
From the above discussion, it follows that the times of the state changes of
a message can be tracked by means of the values stored in the various times-
tamp fields of the corresponding record. For instance, in the case in Fig. 8(c),
we can deduce that the message has been generated on Oct. 16th, 2013
14:15:37.884 (timestamp field), has been waited to be transmitted to the cen-
tral server until 14:17:05.551 of the same day (receipt server timestamp field),
and has been finally delivered to its recipient at 14:21:59.135 (receipt device timestamp
field).

4.2.5 Determining the partners of a message
In addition to user-to-user communication, WhatsApp provides its users with
two forms of collective communications, namely:
• broadcast (i.e, one-to-many) communication, whereby a user (the source

user) sends the same message to a group of other users (the destination
users) that are not aware of each other and whose possible replies are
sent to the source user only;
• group chats, providing a many-to-many communication service, whereby

each message sent by any user belonging to a chat is received by all the
users belonging to that chat.
While the WhatsApp ID of the communication partner in a user-to-user

communication is easily retrieved from the key remote jid field, to determine
the set of users involved into a broadcast or a group chat message various
fields have to be correlated, as discussed below.
Broadcast messages When a user sends a broadcast message, a distinct

record is created in his/her messages table for each one of the recipients,
plus one for itself, as reported in Fig. 9(a), that shows the records generated
by a broadcast message sent to users 39320xxxxxxx, 39335xxxxxxx, and
39333xxxxxxx.
As shown in this figure, all the records corresponding to the same broad-
cast message have the same message identifier (stored in the key id field),
so they can be easily identified. Each one of these records stores in the
key remote jid field the WhatsApp ID of the recipient (the sender uses the
keyword broadcast to denote itself as a recipient), while the remote resource
and the recipient count fields store the WhatsApp IDs of the set of destina-
tions and how many they are, respectively (field needs push instead always
stores the value ’2’).
The situation on each one of the destinations is instead different (Fig. 9(b)),
since each one of them stores, in his/her messages table, only a single record
that is generated when it receives the broadcast message. This record can be
distinguished from those corresponding to non-broadcast messages by look-
ing at the value stored in its key id field, that consists in the concatenation
of the %∼ characters with the message identifier set by the sender.

Figure 9: Records generated for a broadcast message sent to three recipients
on: (a) the sender, (b) one of the recipients. Only the fields that contribute
to the identification of the partners are displayed.
Group chat communication When a message is sent within a group

chat, a record is generated in the messages table of all the members of
that group (including the sender). Each one of these records stores, in the
key remote jid field, the identifier of the group (the group id), a string for-
matted as {creator’s phone number}-{creation time}@g.us (where creation
time is encoded as a Unix epoch time).
To illustrate, consider a group chat consisting of three members, namely
3933xxxxxxx, 3936xxxxxxx, and 3932xxxxxx (in the following denoted as
A, B, and C, respectively, for brevity), where each user sends in turn to the
group a message with textual content ’Message from X’ (where ’X’ is the
name of the user).
Let us focus on the records stored in the messages table of user A at the
end of this exchange, that are shown in Fig. 10 (the situation for the other
users is identical).
Figure 10: Records correspoding to three messages exchanged within a group

chat.

As can be seen from this figure, all these records store the same group id
3933xxxxxxx-1363078943@g.us in the key remote id field. From this value,
we can determine the creator of the group (user A) and the date and hour of
group creation (March 12, 2013 at 09:02:23). Furthermore, the WhatsApp
ID of the message originator is stored into the remote resource field, while the
time of message receipt is stored into the timestamp field. Note that A stores
also the record corresponding to the message that (s)he has sent to the group
(record no. 1 in the figure). Records like this one can be easily identified by
looking at the contents of their status and remote resource fields, that store
the values ’4’ and ’null’, respectively.
Note also that the set of recipients, i.e. of the set of group members at
the time of the sending, is not stored anywhere on the record. However,
it can be indirectly determined by examining the records corresponding to
the control messages that are automatically exchanged by the various group
members each time a user joins or leaves the group. These messages, also
stored in the messages table, always contain the value ’6’ in the status field,
and encode in the media size field the specific operation corresponding to the
message (in particular, the values ’1’, ’4’, and ’5’ indicate group creation,
join, and leave, respectively).
To illustrate, let us consider a scenario in which user 39320xxxxxxx (D,
for brevity) creates a group on Nov. 11, 2013 at 16:24:05, and immediately
adds user 39335xxxxxxx (E, for brevity) to the group. Then, D adds user
39333xxxxxxx (F, for brevity) on Nov. 12, 2013 at 10:40:48.
The records generated by these operations in the chat database of user
D are shown in Fig. 11 (the same situation occurs on all the other group
members).
Figure 11: Group management records stored in the msgstore database of

user D. For other users we have the same situation, with the exception of
record no. 1.
Group creation corresponds to record no. 1, as can be seen from sta-
tus=’6’ and media size=’1’. The time of group creation can be ascertained

(besides from the group id) from the value stored in the timestamp field,
while the field data stores the name given to the group (wa test group).
The addition of user E corresponds instead to record no. 2: the specific
operation (join) and the identity of the user joining the group (E) can be
deduced from fields media size and remote resource field, while the time of
occurrence is stored in the timestamp field. A similar situation occurs with
the addition of user F, whose control message corresponds to record no. 3.
Now, suppose that at a later time, namely on Nov. 14, 2013 at 22:11:36,
user F leaves the group. This operation corresponds to record no. 4 in
Fig. 11 (media size=’5’ indicates a group leave), that reports the identity of
the user leaving the group and the time of leave in the remote resource and
the timestamp field, respectively. Finally, when user E leaves the group on
Nov. 15, 2013 at 09:49:54, record no. 5 is added to the messages table.
By using the information discussed above, the composition of the group
over time can be reconstructed by chronologically sorting the various con-
trol messages corresponding to join (status=’6’ and media size=’4’) and leave
(status=’6’ and media size=’5’) of a given group (identified by the contents
of the key remote jid field), as shown in Fig. 12.
Figure 12: Timeline of group composition variations.
From this information, it can be inferred whether a user belonged or not

to a group when a specific message was sent to that group.
4.2.6 Dealing with deleted messages

In WhatsApp Messenger, the user may delete the records stored in the msg-
store.db database in two different ways, namely:
• deletion of an individual message: in this case, the corresponding record

is deleted from the messages table;

• deletion of all the records belonging to a one-to-one, broadcast, or
group chat conversation: in this case, all the records corresponding
to the messages exchanged in that conversation are deleted from the
messages table, as well as the record of the chat list table corresponding
to that conversation.
As discussed before, it is sometimes possible to recover deleted SQLite

records, and in these cases the analysis techniques discussed in the previous
sections can be applied.
However, when such a recovery is not feasible, it may be still possible to
determine many of the information regarding a deleted message by analyzing
the log files generated by WhatsApp Messenger. In particular, as discussed
below, it is possible to determine which messages have been deleted and
when, when a deleted message has been sent or received and its state, as
well as the users involved in the conversation. The same holds true for group
control messages, so the analysis of log files also makes it possible to track
the evolution of each group over time. In other words, only the contents of
a deleted message cannot be recovered anymore.
Finding deleted messages and their deletion times When a message

is deleted, WhatsApp Messenger records into the log file an event like the one
shown in Fig. 13, that indicates both the type of operation (msgstore/delete)
and the identifier of the deleted message (1363253484-1), as well as the time
of deletion (March 14, 2013 at 10:49:22).
Figure 13: Events logged when a message is deleted.
Determining when a deleted message has been exchanged, and

its state Each time a user-to-user, broadcast, or group chat message is
sent/received, WhatsApp Messenger logs the time of the send/receive oper-
ation, the identifiers of the involved users, and the identifier of the message.
Therefore, by searching into the log file the events corresponding to exchanges
of deleted messages, it is possible to ascertain when those messages have been
sent or received.

For instance, the event reported in Fig. 14 line 1 corresponds to the send-
ing of the deleted message identified by 1363253484-1 to user 39366xxxxxxx
on March 14, 2013 at 09:37:44.
Figure 14: Events logged when a message is sent.
Finally, WhatsApp Messenger logs also the events corresponding to re-

ception of the acknowledgment messages sent back by the central server (line
no. 5) and by the recipient (line no. 7), from which it is possible to determine
the state of a message, as well as the times of its state changes.
Temporal evolution of group chat composition The evolution of the

composition of a group chat can be tracked over time by examining the
events, logged by WhatsApp Messenger, corresponding to the exchange of
the various control messages discussed in Sec. 4.2.5.
The events corresponding to the creation of a group and to the addition
of two users are shown in Fig. 15 (that reports an excerpt of the log of the
group creator).
The creation of the group gives rise to the events reported in lines no. 1
and 4, from which we can obtain the group name and creation time. The
request to add to the group users 39366xxxxxxx and 39320xxxxxxx corre-
sponds to line 7, while the addition of the former and the latter user corre-
sponds to lines 10–14 and 18–22. Finally, the leave of a user corresponds to
the event logged on line no. 27.
4.3 Analysis of settings and preferences

WhatsApp Messenger stores various information of potential evidentiary
value in several files, located in the directories listed in Table 1 row no.
9.

Figure 15: Events corresponding to group operations.

In particular, the file me stores (as ASCII text) the phone number reg-
istered with WhatsApp (i.e., the number used to create the corresponding
WhatsApp ID). The relevance of this information derives from the fact that
the SIM card currently used with the smartphone may not be the one used
to register the user with WhatsApp: it is indeed possible to replace the latter
SIM card with a new one, and to use the WhatsApp ID corresponding to
the phone number of the old SIM card. Thus, a user A may impersonate a
different user B, as long as A has used B’s SIM card during registration, or
(s)he is using B’s smartphone with a different SIM card. By comparing the
phone number of the SIM inserted into a smartphone with the phone number
stored in the me file, it is possible to determine whether this is the case or
not.
Furthermore, the file me.jpg stores the currently-used avatar picture of the
user. Given that the avatar pictures of all contacts are downloaded locally
by WhatsApp Messenger (as discussed in Sec. 4.1.1), the me.jpg file can be
used to understand that the user of the device under examination has been in
contact with another user even if the latter one has deleted from its contacts
database the record corresponding to the former one. As a matter of fact, the
deletion of a record from the contacts database does not cause the deletion
of his/her downloaded avatar picture.
5 Conclusions
In this paper we have discussed the forensic analysis of the artifacts left
by WhatsApp Messenger on Android smartphones, and we have shown how
these artifacts can provide many information of evidentiary value. In partic-
ular, we have shown how to interpret the data stored into the contacts and
chat databases in order to reconstruct the list of contacts and the chronology
of the messages that have been exchanged by users.
More importantly, we have also shown the importance of correlating
among them the artifacts generated by WhatsApp Messenger in order to
gather information that cannot be inferred by examining them in isolation.
As a matter of fact, while the analysis of the contacts database makes it
possible to reconstruct the list of contacts, the correlation with the events
stored into the log files maintained by WhatsApp Messenger allows the in-
vestigator to infer also when a specific contact has been added, or to recover
deleted contacts and their time of deletion. Similarly, the correlation of the

contents of the chat database with the information stored into the log files
allows the investigator to determine which messages have been deleted, when
these messages have been exchanged, and the users that exchanged them.
The results reported in this paper have a two-fold value. On the one
hand, they provide analysts with the full picture concerning the decoding,
interpretation, and correlation of WhatsApp Messenger artifacts on Android
devices. On the other hand, they represent a benchmark against which the
ability of extraction tools for smartphone to retrieve all the WhatsApp Mes-
senger artifacts can be assessed.
It is however worth to point out that the results discussed in this paper
apply to Android only: as a matter of fact, there is evidence [21] showing that
on different platforms (e.g., iOS) WhatsApp Messenger produces artifacts
that differ either in the information they store, or in their format, or in
both. We leave the analysis of WhatsApp Messenger for other smartphone
platforms as future work.
References
[1] AccessData Corporation. FTK Imager, 2013. Available at
http://www.accessdata.com/support/product-downloads.
[2] N.B. Al Barghuthi and H. Said. Social Networks IM Forensics: Encryp-

tion Analysis. Journal of Communications, 8(11), Nov. 2013.
[3] Steven M. Bellovin, Matt Blaze, Sandy Clark, and Susan Landau. Law-
ful Hacking: Using Existing Vulnerabilities for Wiretapping on the Inter-
net. In Proc. of Privacy Legal Scholars Conference, June 2013. Available
at http://dx.doi.org/10.2139/ssrn.2312107.
[4] Cellebrite LTD. Cellebrite Android Forensics, 2013. Available

at http://www.cellebrite.com/mobile-forensics/capabilities/android-
forensics.
[5] D. Cortjens, A. Spruyt, and W.F.C. Wieringa. What-

sApp Database Encryption Project Report. Technical re-
port, 2011. Available at https://www.os3.nl/ media/2011-
2012/students/ssn project report.pdf.

[6] Don Ho. Notepad++ Home, 2013. Available at http://notepad-plus-
plus.org.
[7] Andrew Hoog. Android Forensics: Investigation, Analysis and Mobile
Security for Google Android. Elsevier - Syngress, 2011.
[8] MohammadIftekhar Husain and Ramalingam Sridhar. iForensics:
Forensic Analysis of Instant Messaging on Smart Phones. In Sanjay
Goel, editor, Digital Forensics and Cyber Crime, volume 31 of Lecture
Notes of the Institute for Computer Sciences, Social Informatics and
Telecommunications Engineering. Springer Berlin Heidelberg, 2010.
[9] S. Jeon, J. Bang, K. Byun, and S. Lee. A recovery method of deleted
records for SQLite database. Personal and Ubiquotous Computing, 16,
2012.
[10] A. Mahajan, M.S. Dahiya, and H.P. Sanghvi. Forensic Analysis of In-
stant Messenger Applications on Android Devices. International Journal
of Computer Applications, 68(8), April 2013.
[11] Micro Systemation. XRY, 2013. Available at
http://www.msab.com/xry/xry-current-version.
[12] Nathan Olivarez-Giles. Whatsapp adds voice messaging as it hits
300 million monthly active users. The Verge, Aug 2013. Available
at http://www.theverge.com/2013/8/6/4595496/whatsapp-300-million-
active-users-voice-messaging-update.
[13] Oracle Corp. Oracle VM Virtual Box. http://www.virtualbox.org, 2013.
[14] Oxygen Forensics, Inc. Oxygen Forensics, 2013. Available at
http://www.oxygen-forensic.com/en/features/analyst.
[15] Oxygen Forensics, Inc. SQLite Viewer, 2013. Available
at http://www.oxygen-forensic.com/en/features/analyst/data-
viewers/sqlite-viewer.
[16] Ivo Pooters, Pascal Arends, and Steffen Moorrees. Extract-
ing SQLite records: Carving, parsing and matching. Tech-
nical report, Digital Forensics Research Workshop Chal-
lenge, 2011. Available at http://sandbox.dfrws.org/2011/fox-
it/DFRWS2011 results/Report/Sqlite carving extractAndroidData.pdf.

[17] Fabio Sangiacomo and Martina Weidner. What-
sApp Xtract (v. 2.1), May 2012. Available at
https://code.google.com/p/hotoloti/downloads/list.
[18] SQLite Consortium. SQLite Home Page, 2013. Available at

http://www.sqlite.org.
[19] N.S. Thakur. Forensic Analysis of WhatsApp on Android Smartphones.

Master’s thesis, University of New Orleans, 2013. Paper 1706.
[20] The United Nations Office on Drugs and Crime. Comprehensive

Study on Cybercrime. Technical report, United Nations, Feb.
2013. Available at http://www.unodc.org/documents/organized-
crime/UNODC CCPCJ EG.4 2013/CYBERCRIME STUDY/
210213.pdf.
[21] Yu-Cheng Tso, Shiuh-Jeng Wang, Cheng-Ta Huang, and Wei-Jen Wang.
iPhone Social Networking for Evidence Investigations Using iTunes
Forensics. In Proceedings of the 6th International Conference on Ubiqui-
tous Information Management and Communication, ICUIMC ’12, New
York, NY, USA, 2012. ACM.
[22] Petr Vanek and Kamil Les. Sqlite Databases Made Easy, 2013. Available
at http://sqliteman.com.
[23] Timothy Vidas, Chengye Zhang, and Nicolas Christin. Towards a Gen-
eral Collection Methodology for Android Devices. Digital Investigation,
8, Aug 2011.
[24] WhatsApp Inc. WhatsApp, 2013. Available at

http://www.whatsapp.com.
[25] Rolfe Winkler. WhatsApp Hits 400 Million Users, Wants to Stay In-
dependent. The Wall Street Journal - Digits, Oct. 2013. Available
at http://blogs.wsj.com/digits/2013/12/19/whatsapp-hits-400-million-
users-wants-to-stay-independent.
[26] YouWave Corp. Youwave home page, 2013. Available at

http://youwave.com.


Forensic Analysis of Whatsapp Messenger On Android Smartphones

Uploaded by

Copyright:

Available Formats

Forensic Analysis of Whatsapp Messenger On Android Smartphones

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

Forensic Analysis of Whatsapp Messenger On Android Smartphones

Uploaded by

Copyright:

Available Formats

Forensic Analysis of WhatsApp Messenger

This manuscript is the preprint of the paper

2 available under the CC-BY-NC-ND

3 available under the CC-BY-NC-ND

3 Analysis methodology and tools

4 available under the CC-BY-NC-ND

• they are device-specific: this would prevent a third party to replicate

For these reasons, in our study we adopted a different approach, in which

5 available under the CC-BY-NC-ND

4 Forensic analysis of WhatsApp Messenger

6 available under the CC-BY-NC-ND

Table 1: WhatsApp Messenger artifacts.

7 available under the CC-BY-NC-ND

4.1 Analysis of contact information

4.1.1 Retrieving contact information

8 available under the CC-BY-NC-ND

Table 2: Structure of the wa contacts table

9 available under the CC-BY-NC-ND

4.1.2 Determining when a contact has been added

4.1.3 Dealing with blocked contacts

10 available under the CC-BY-NC-ND

The information concerning blocked users is stored neither in the contacts

11 available under the CC-BY-NC-ND

4.1.4 Dealing with deleted contacts

12 available under the CC-BY-NC-ND

4.2 Analysis of exchanged messages

4.2.1 The structure of the chat database

13 available under the CC-BY-NC-ND

Table 3: Structure of the messages table: fields storing message attributes.

14 available under the CC-BY-NC-ND

Table 4: Structure of the messages table: fields storing information concern-

Field name Meaning

Table 5: Structure of the chat list table.

15 available under the CC-BY-NC-ND

4.2.2 Reconstruction of the chat history

Figure 3: Reconstruction of the chat history. Phone numbers have been

16 available under the CC-BY-NC-ND

4.2.3 Extracting the contents of a message

17 available under the CC-BY-NC-ND

On the recipient side, after message reception, the transmitted thumbnail

Figure 5: Multimedia file exchange: recipient side

18 available under the CC-BY-NC-ND

Figure 6: Fields containing contact card information.

Geolocation coordinates WhatsApp Messenger enables users to send the

4.2.4 Determining the state of the message

19 available under the CC-BY-NC-ND

20 available under the CC-BY-NC-ND

21 available under the CC-BY-NC-ND

• broadcast (i.e, one-to-many) communication, whereby a user (the source

• group chats, providing a many-to-many communication service, whereby

While the WhatsApp ID of the communication partner in a user-to-user

Broadcast messages When a user sends a broadcast message, a distinct

22 available under the CC-BY-NC-ND

Group chat communication When a message is sent within a group

Figure 10: Records correspoding to three messages exchanged within a group

23 available under the CC-BY-NC-ND

Figure 11: Group management records stored in the msgstore database of

24 available under the CC-BY-NC-ND

Figure 12: Timeline of group composition variations.