An IAC Approach for Detecting Profile Cloning in Online Social Networks

International Journal of Network Security & Its Applications (IJNSA), Vol.6, No.1, January 2014
DOI : 10.5121/ijnsa.2014.6107 75
An IAC Approach for Detecting Profile Cloning
in Online Social Networks
MortezaYousefi Kharaji1
and FatemehSalehi Rizi2
1
Deptartment of Computer and Information Technology Engineering,Mazandaran
University of Science and Technology,Babol, Iran
2
Department of Computer Engineering and Information Technology, Sheikhbahaee
University of Isfahan, Isfahan, Iran
Abstract
Nowadays, Online Social Networks (OSNs) are popular websites on the internet, which millions of users
register on and share their own personal information with others. Privacy threats and disclosing personal
information are the most important concerns of OSNs’ users. Recently, a new attack which is named
Identity Cloned Attack is detected on OSNs. In this attack the attacker tries to make a fake identity of a real
user in order to access to private information of the users’ friends which they do not publish on the public
profiles. In today OSNs, there are some verification services, but they are not active services and they are
useful for users who are familiar with online identity issues. In this paper, Identity cloned attacks are
explained in more details and a new and precise method to detect profile cloning in online social networks
is proposed. In this method, first, the social network is shown in a form of graph, then, according to
similarities among users, this graph is divided into smaller communities. Afterwards, all of the similar
profiles to the real profile are gathered (from the same community), then strength of relationship (among
all selected profiles and the real profile) is calculated, and those which have the less strength of
relationship will be verified by mutual friend system. In this study, in order to evaluate the effectiveness of
proposed method, all steps are applied on a dataset of Facebook, and finally this work is compared with
two previous works by applying them on the dataset.
Keywords
Online social networks, Profile cloning, Privacy
1.Introduction
Social network websites are defined as web services that allow users to make public and semi-
public profiles in a bounded system, to build a list of users with whom have a kind of common
relationship, and to search in their friends’ lists [1]. One of the most important challenges of
observing friends’ information is threatening users’ security and privacy. An adversary can cause
many problems by exploiting users’ information. This data may contain users’ financial
information which adversary can use them to do identity theft attacks, or may contain users’
medical background such as healthy status, diagnosis or treatment records [2].
Recently, a new kind of attack which is named Identity Clone Attack is detected on OSNs that
makes fake identities of specific users. The basic goals of the adversary in this attack are
obtaining victim’s friends’ personal information by forging real user profile, and increasing trust
among mutual friends to do more defrauding in the future [3]. Two kinds of these attacks are
already defined: first one is Single-Site Profile Cloning, and the next one is Cross-Site Profile
Cloning. In the first attack, adversary forges the real user profile in the same social network and
use this cloned profile to send friend request to users’ friends. An unaware user may think this

76
request is from a familiar user hence she/he will confirm it and his/her personal information will
be accessible for adversary. The next attack is cross-site profile cloning, as it shown in Figure1,
the adversary detects a user with his/her friends in network A, then make a clone profile with
his/her attributes in network B which user has not made account yet. The adversary sends friend
requests to the victim’s friends in network B. Victim’s friends think they know the sender of
requests and confirm them, and as soon as they confirm the request, the adversary will thieve their
personal information. The adversary uses this information to make other clone profiles or to
deceive others in the future. Detecting this kind of attack is very difficult for service providers
and profiles owners, because service providers think it is a new user which is registering in these
websites [4]. Discovering cloned profiles with more precise methods can bring more security for
users who are using social networks, and also cause an increasing movement for service providers
to improve their security level in the services they provide on their platforms [5].
Figure 1. Single-site profile cloning and cross-site profile cloning attacks [5]
The rest of the paper is organized as follows: in section 2,a short review on related works are
expressed and section 3 presents the proposed method for detecting cloned profiles in complete
details. In section 4, to evaluate the applicability of proposed method, it is applied on a dataset of
Facebook and it is also compared with previous works in section 5. Finally, in section 6, the paper
is concluded and some feasible future works are discussed.
2.Related Works
Many social networks have a weak user to user authentication mechanism that are mostly based
on presented information such as name, photos, and a set of social links. This causes the misuse
of profile cloning attack to make fake social links. Bhumiratana in [6] presented a model to
exploit of available weak trust in social networks. This model saves the authority of an online
fake identity which made by profile cloning attack to obtain more personal information. This
research proposed an attack methodology to use cloned profiles and to do reliable interactions
among selected users. Proposed model uses an array of attacking techniques to make a permanent
and automatic cloned identity of real users on social networks so that are able to get personal data
in a specific period of time. This proposed system works among different social networks.
Jin et al. in [7] proposed an active detection framework to detect cloned profiles.An intelligent
fake identity not only forges users' attributes, but may add victim's friends into his friend network
too. According to similarity of attributes and users' friend list there are two ways for defining
similarity measure among real identity and fake identities. One of them is basic profile similarity
and the next one is multiple-faked identities profile similarity. In this research, according to the
similarity of profiles, a framework for detecting cloned profiles on social network is proposed
which contains of three steps: first step is to search and separate identities as a set of profiles, as

77
the entry of search is a profile attributes. Second step is detecting suspicious profiles by using
profile similarity schemas, and third step is deleting cloned profiles from friend list. In detecting
process adjusting a set of parameters can help to do a correct detection in different social
networks.
Kontaxis et al. in [8] offered a tool which is able to automatically search and detect cloned
profiles in OSNs. The concept key of their approach is using user-specific data which is extracted
from real user profile in social network. In this approach, finally a list of profiles which are
probably cloned with similarity scores is presented to user. A string matching algorithm is used to
define the similarity of attributes between two profiles and assign similarity score for each
candidate identity. In this method detecting cloned profile contains three steps as follows:
information Distiller, profile hunter, and profile verifier.
Gani et al. in [9] discussed a piece of work which intends to provide some insights regarding the
resolution of the hard problem of multiple identities detection. Based on hypothesis that each
person is unique and identifiable whether in its writing style or social behavior, they proposed a
framework relying on machine learning models and a deep analysis of social interactions, towards
such detection.
Most of the current research has focused on protecting theprivacy of an existing online profile in a
given OSN. Instead, Conti et al. in [10]noted that there is a risk of not having a profile in the last
fancysocial network. The risk is due to the fact that an adversary maycreate a fake profile to
impersonate a real person on the OSN.The fake profile could be exploited to build online
relationshipwith the friends of victim of identity theft, with the final target ofstealing personal
information of the victim, via interacting onlinewith the friends of the victim.
3. The proposed approach
The detection approach is organized in 6 steps as follows:
3.1. Discovering community the social network graph
In many social networking sites, network topological structure and attributes values are the
complete information. Nodes represent users and edges represent the relationship among them. In
each node, there are some attributes such as name, gender, education, interests, location and
social activities. It is obvious that network topological structure and attribute information can be
used to identify some hidden patterns in communities. In this study, IAC clustering algorithm [11]
is applied to detect communities in social network graphs. Figure 2 shows a pseudo code of the
algorithm where it accepts an attribute augmented graph and return a clustered graph as output.
Figure 2. IAC Clustering Algorithm [11]

78
An augmented graph is a graph G = (V, E, ), where V = {v1, v2, v3, …,vn } is the set of nodes
and n = |V| denotes the number of nodes in the graph, E ⊂ V× V is the set of edges,E = {(vi, vj):
vi, vj∈V}, and ∈ R|v| × d
is the nodes attribute matrix. First of all, the algorithm creates the
similarity matrix C, then according to K (K = × E) it adds the set of edges to the graph and the
elements which belong to these edges are set to 1 in matrix S. As well as matrix W is made by
summation of S and A. To this end, a weighted graph is clustered by MCL algorithm that is
demonstrated in Figure 3. MCL is a clustering algorithm [12] based on stochastic flows on the
graph and in order to execute it, first, transition matrix should be made from weighted graph
obtained through matrix W. This algorithm includes expansion and inflationoperations on
stochastic matrixes such that the expansion is calculated as M×M and the inflation increases the
M’s elements to amount of r (r 1), then normalizes each column. Eq. 1 indicates how the
inflation operation works, after normalizing the summation of each column will be 1.
ΓM

(1)
MCL is started from a standard flow matrix and the two operations apply it alternatively until the
output matrix gets a stable state and it will not be changed when the operations are applied again.
After,allof clusters are determined in the rows of the stable matrix.
Figure 3. MCL clustering algorithm [12]
3.2.Extraction user’s attribute
In this stage, the user’s information is extracted from his/her legitimate profile in online social
network. At the start, the user’s profile is analyzed then it is specified that which parts of user’s
profile can be regard as user-specific. This information is used to construct queries in search
engines of social networks. The extracted information is includes name, gender, location,
education, email and etc. social networks owner and service provider have complete access to
users’ data and can exploit user-specific from her/his profile easily.
3.3.Search in community
In step1, the socialgraphwasclusteredconcerning to users’ attribute similarities. In this stage for
finding similar profiles to real user’s profile, the cluster which is belong to real user is marked
then all of similar profiles are searched by name attribute. The search result is the list of profiles
with similar or same name to real profile.

International Journal of Network Security Its Applications (IJNSA), Vol.6, No.1, January 2014
79
3.4. Selecting profile
In this stage, the profiles which have mutual friends with victim (real profile) are picked up
among founded profiles in step 3. Mutual friends are the friends who exist in the victim’s friend
list and in the friend list of each candidate profile in the same time. Since, in profile cloning
attacks many friend requests are sent to victim’s friends, it is obvious they have some common
friends with victim [4]. Hence, only profiles which have mutual friends with victim are chosen
for continuing next steps.
3.5. Computing strength of relationship
In step 5, all of nodes’ edges which was acceded in this stage, are weighted considering to the
number of common active friends, shared Urls and page-likes among users. Formally, the social
network can be defined as a weighted graph G = (V, E, W), where V is the set of profiles, E ⊆ V
× V is the set of edges, and W ⊆ℜ is a set of weights are assigned to edges. For each node v ∈ V,
a 3-dimentional feature vector is defined as it is included in the number of active friends, page
likes and common shared URLs. Therefore, weight of each edge eij = (vi ,vj) is calculated as
summation of common actives friends, page likes and common shared URLs between nodes vi
and vj. Further details presented how the weights can compute come in the following parts [13].
3.5.1 Active friends:
This measure takes the interaction frequency of a user with his/her friends in the network. For a
user Vi with Fi as the set of friends, the set of active friends Fi
a
can be computed as an interaction
between the set Fi and the set of friends of Vi who were either contacted by Vi or those who
interacted with Vi through wall posts, comments or tags. It can be defined using Eq. 2 in where Ii
is the set of users with whom Vihas interactions in the network. For a node Vi the value of the
“active friends” feature is taken as the cardinality of the set of its active friends Fi
a
. Similarly, the
set of common active friends in the network with whom a pair of users vi and vj have interacted is
calculated as the intersection of their active friends Fi
a
and Fj
a
, respectively, as given in Eq. 3. For
an edge eij = (vi ,vj), the value of the “active friends” feature is taken as the cardinality of the set
of common active friends Fij
a
[13].

(2)

(3)
3.5.2 Pages-likes:
This feature computes the page likes frequency of the users in social network. For an edge eij = (vi
, vj), the common page likes of vi and vj, Pij, is calculated as the interaction of the sets of page
likes of vi and vj, as given in Eq. 4, and the page likes attribute value is calculated as the
cardinality of the set Pij[13].

(4)
3.5.3 URLs:
this feature captures the URL sharing patterns of the social networks users. For an edge eij = (vi
,vj), the common URLs of vi and vj, Uij, is calculated as the intersection of the set of URLs shared
by vi and vj. The URLs attribute value is calculated as a fraction of URLs commonly shared by
them using Eq. 5 [13].

(5)

80
On the basis of the above mentioned features, each edge eij = (vi ,vj), is assigned a weight w(eij)
that is calculated as an summation of the individual feature value as given in Eq. 6. ||represents
the cardinality of the set [13].
! |

| || || (6)
Afterward, the weights are assigned to each edge in social network graph and strength of
relationship is calculated between two nodes as follows:
3.5.4 Definition 1 (Friendship Graph) [14]
Given a social network G and a node v ∈ G.N, the friendship graph of v, denoted as FG (v), is a
sub-graph of G where: (1) FG(v).N = {v} {n ∈ G.N | n ≠ v, ∃ e ∈ G.E, e = v, n }; (2)
FG(v).E = {e = v, n∈ G.E | n ∈ FG (v).N} {e = n, n′∈ G.E | n, n′ ∈ FG(v).N }
3.5.5 Definition 2 (Mutual Friends Graph) [14]
Given a social network G and two nodes v, c ∈ G.N, the mutual friends Graph of v and c, denoted
as MFG(v, c), is a sub-graph of G where: (1) MFG(v, c).N = {v, c} {n ∈ G.N | n ≠ v, n ≠ c, ∃ e,
e′∈ G.E, e = v, n∧ e = n, c }; (2) MFG(v, c).E = {e = n, n′ ∈ G.E | n, n′ ∈ MFG (v, c).N}
For instance, Friendship graph of node 7 and mutual friends graph of 7 and 12 are shown in
Figure 4.
Figure 4. Friends and mutual friends graphs
3.5.6 Definition 3 (Strength of relationship between two nodes)
Given a social network G and two nodes v, c ∈ G.N, Let T = {MFG(v, c).E }, R = { FG(v).E }, P
= {FG (c).E }. Strength of relationship between v and c is defined in Eq.7 as follows:
'(), +,
∑ -.
.∈/
∑ -.
.∈0 1 ∑ -.2
.2∈3
(7)
Strength of relationship (SR)measure is calculated between each suspiciousprofile which
hasmutual friends with victim. Inasmuch as an expert adversary attempts to make less suspicious
by making social relationship and interactions with victim’s friends. Strength of relationship
measure is used to detect cloned identities because the real identities make more deep social
activities than them as they mostly know each other in real life. They might get intimacy through
relationships in real life or voice and video chat on the Internet for a while [15]. Therefore, real

81
users contribute in social activities like commenting, sending message and tagging more than
fakes and clearly they have higher SR comparing to cloned profiles. In the rest of this stage,
nodes are sorted in a list by amount of SR as RS (v, c1) RS (v, c2)RS (v, c3) …RS (v, cn)
and n is the number of profiles which have reached in step 5. Among these profiles, c1 has the
least SR and it will be sent to next step for verifying. If it does not identify as a cloned identity the
next one in the list, c2 will be gone to stage 6. This trend will be continued until the last profile in
the list.
3.6. Decision making
Heretofore, some methods were presented to verify the suspicious identity in online social
networks. In a primary approach, the ID number is asked from users for verification process. For
example Identity Badge wants users to enter their passport number [16]. The social verification
approach is presented by Schechter et al. [17] want users to design some questions to verify their
friends and if a user answers most questions correctly he/she will be marked as a valid user. A
proposed approach is verifying suspicious identities by mutual friends as it is wanted mutual
friends to design some question concerning to background knowledge that they have obtained
during their relationships. As well as these questions can be design by some social engineering
teachings. It is evident that a cloned identity cannot answer the question correctly, specially the
questions which are designed considering to users’ background. Also similar identities (are not
fake) send their own answers which are understandable for mutual friends as they come from real
identities. Eventually, fake identities are identified and they remove or closed temporary by
service provider as well as their friends receive some notification for existence a fake identity in
their friend list. Figure 5demonstrates a view of proposed verification system.
Figure 5. Verification system by mutual friends
The diagram of detection approach is shown in Figure 6.
Figure 6. IAC Detection Approach

82
4.Experimental Results
In order to evaluate the proposed approach, an office dataset of Facebook users [18] is used and it
is updated by adding user’s attributes, shared Urls and page-likes. Verification the proposed
approach is not possible for a normal social network user because only service providers
haveaccess to users’ original information and social network graph.Also some social networks
have restrictions thus normal users cannot make clone profile easily [4]. There are 63,731 users in
this dataset and 1,634,115 links among them thus each user has 25.6 relationship links on
average. To evaluate the approach, it is assumed that there are some fake identities in this dataset
and it is necessary to add themto dataset as victims. For demonstrating the detail of effectiveness
of proposed approach, 20 users are selected from the dataset as their social graph is shown in
Figure7 and their attributes are exhibited in Table 1.
Table 1. 20 users’ attributes selected from dataset
ID Name
Gende
r
Education Work
Birthd
ay
Locatio
n
Relatio
nship
School Degree
Employe
r
Position
32
NikoP
arda
Femal
e
Harvard University PhD East Man Manager 1979 USA Single
35
Sara
Abraha
m
Femal
e
Arcadia University Master’s Owens Web Developer 1980 USA Single
36
Sara
Abraha
Femal
e
Carolina University Master’s Owens Web Developer 1980 USA Single
174
David
Ernox
Male Michigan University Master’s Qpass Java Developer 1984 USA Single
463
Sara
Abram
Femal
e
Michigan University Master’s AppNet Web Developer 1985 USA Single
1236
Tom
Banho
Male Acaedia University Bachelor Xing
Network
Manager
1979 USA Married
2411
Rose
Milan
Femal
e
Koln University PhD Axvert Manager 1972 USA Single
33
Hanrry
Dabuo
Male Dublin High school Diploma Sonic Secretary 1970 UK Married
34
Rosa
Morad
a
Femal
e
Franklin High
school
Diploma Sonic Bookkeeping 1974 UK Married
163
Charls
Selvin
Male
Pietersburg
University
Bachelor Sony Accountant 1979 UK Married
4013
SeolDi
ao
Male Chester University Master’s Maxtor
Database
Administrator
1983 France Single
4014
Lore
Parsan
Femal
e
Pietersburg
University
Bachelor Sonic
Database
Administrator
1982 Spain Single
4023
Caroli
n Wolf
Femal
e
Franklin High
school
Diploma Sony Bookkeeping 1979
German
y
Married
1081
Alex
Monat
a
Male Lowa University Master’s Sony
Electrical
Engineer
1986 UK Married

83
37
Silvia
Jacson
Femal
e
Carolina University Bachelor MySpace
Computer Data
Clerk
1978
Australi
a
Married
1187
Shery
Monat
en
Femal
e
Dublin High school Diploma MySpace Buyer 1968
Australi
a
Single
1195
Melina
Diyana
Femal
e
Pietersburg
University
PhD MySpace
Call Center
Assistant
1989
Australi
a
Single
1234
LinaEg
hose
Femal
e
Gabelino High
school
Diploma Amgen Buyer 1980 Canada Single
1235
Mariya
naPlan
ta
Femal
e
lowa University Bachelor Amgen
Electrical
Engineer
1987 Canada Single
1237
Toney
Cazola
Male Carolina University Bachelor Amgen
Call center
Operator
1978 Canada Single
Figure 7.20 users’ social graph
As mentioned before, an active friend is a friend who posts on the wall, comments and tags on
her/his friends’ posts. This relationship is shown in green lines in Figure7.
4.1. Testing the IAC approach on dataset
All of detection steps (6 steps) are applied to users of dataset as well as it is supposed that they do
not use any particular privacy setting.
4.1.1.Choosing a victim identity
Initially, a user is selected as a victim identity from dataset. As it is mentioned in section 1, an
attacker makes a fake identity considering some acceptable information of a real identity which
he/she has already gathered from online social networks or other sites. Attacker uses this victim to
reach his goal by connecting to victim’s friends [4].User 35 is chosen as a victim because it has
some perquisites as the number of links (edges) and social activities (green edges) in the network.
Therefore a victim identity 35′ is created and its attribute values are displayed in Table 2 and
Figure 8 demonstrates its position in social graph in red color.

84
Table 2. 20 users’ attributes selected from dataset with fake identity
ID Name Gender
Education Work
Birthday
Locati
on
Relations
hip
School Degree Employer Position
32 NikoParda Female
Harvard
University
PhD East Man Manager 1979 USA Single
35
Sara
Abraham
Female
Arcadia
University
Master’s Owens Web Developer 1980 USA Single
35′
Sara
Abraham
Female
Arcadia
University
Bachelor Owens Web Developer 1980 USA Single
36 Sara Abraha Female
Carolina
University
Master’s Owens Web Developer 1980 USA Single
174 David Ernox Male
Michigan
University
Master’s Qpass Java Developer 1984 USA Single
463 Sara Abram Female
Michigan
University
Master’s AppNet Web Developer 1985 USA Single
1236 Tom Banho Male
Acaedia
University
Bachelor Xing
Network
Manager
1979 USA Married
2411 Rose Milan Female Koln University PhD Axvert Manager 1972 USA Single
33
HanrryDabu
o
Male
Dublin High
school
Diploma Sonic Secretary 1970 UK Married
34
Rosa
Morada
Female
Franklin High
school
Diploma Sonic Bookkeeping 1974 UK Married
163 CharlsSelvin Male
Pietersburg
University
Bachelor Sony Accountant 1979 UK Married
4013 SeolDiao Male
Chester
University
Master’s Maxtor
Database
Administrator
1983 France Single
4014 Lore Parsan Female
Pietersburg
University
Bachelor Sonic
Database
Administrator
1982 Spain Single
4023 Carolin Wolf Female
Franklin High
school
Diploma Sony Bookkeeping 1979
Germa
ny
Married
1081 Alex Monata Male
Lowa
University
Master’s Sony
Electrical
Engineer
1986 UK Married
37 Silvia Jacson Female
Carolina
University
Bachelor MySpace
Computer Data
Clerk
1978
Austral
ia
Married
1187
SheryMonat
en
Female
Dublin High
school
Diploma MySpace Buyer 1968
Austral
ia
Single
1195
Melina
Diyana
Female
Pietersburg
University
PhD MySpace
Call Center
Assistant
1989
Austral
ia
Single
1234 LinaEghose Female
Gabelino High
school
Diploma Amgen Buyer 1980
Canad
a
Single
1235
MariyanaPla
nta
Female lowa University Bachelor Amgen
Electrical
Engineer
1987
Canad
a
Single
1237
Toney
Cazola
Male
Carolina
University
Bachelor Amgen
Call center
Operator
1978
Canad
a
Single

85
Figure 8. 20 users’ social graph with a fake identity
4.1.2. Initializing

As mentioned in section 3-1, it is necessary to initialize before performing the experiments on
thedataset. Attribute augmented edges are chosen among the top K similar pairs of matrix C
where K = × |E|. The higher mount is gotten, the more edges are added to each community
thus more accurate clusters are formed on the social graph. At the beginning, is set by 0.68 (K =
34) then it will be set by other values in section 4-2.
4.1.3.Discovering communities in social graph
After performing IAC algorithm on dataset, the attribute augmented graph and clustered graph
with three communities C1, C2 and C3are gained so that isshown in Figure 9 and Figure 10
respectively.
Figure 9. Attribute augmented graph

86
Figure 10.Clustered graph by IAC algorithm
4.1.4. Extracting victim’s attributes
The information of victim (who wants to detect his clones) is extracted in this step and it shown in
Table 3.
Table 3. Real user’s attributes
ID Name Gender
Education Work Birthda
y
Location
Relationshi
p
35
Sara
Abraham
Female
Arcadia
University
Master’s Owens
Web
Developer
1980 USA Single
4.1.5.Searching in Community
Since node 35 is belong to C1, only in this community is search for finding similar profiles to 35.
The searchresult is shown in Table 4.
Table 4.Similar users to 35
ID Name Gender
Education Work Birthd
ay
Location Relationship
35′
Sara
Abraha
m
Female
Arcadia
University
Bachelor Owens
Web
Developer
1980 USA Single
36
Sara
Abraha
Female
Carolina
University
Master’s Owens
Web
Developer
1980 USA Single
463
Sara
Abram
Female
Michigan
University
Master’s AppNet
Web
Developer
1985 USA Single

87
4.1.6. Selecting apt identities
According to profile cloning attacks, an attacker aims victim’s friends and sends them friend
requests hence a cloned profile will have some victim’s friends in its friend list [29]. Node 463 is
not a clone identity because it is connected to node 35 directly and only 36 and 35′ are passed to
next step.
4.1.7. Computing strength of relationship
In this step, SR is calculated for node 35′ and node 36 in regard to Eq. 2,3,4,5,6,7 then they will
be ordered by values:
SR (35, 35′) = 14.497
SR (35, 36) = 36.85
As it is shown amount of SR (35, 35′) is less than other and first it will be sent to next stage for
verification.
4.1.7. Verification
In this part, nodes 2411, 32, 1236, 174 (mutual friends between 35 and 35′) are asked to design
some technical questions concerning the relationship background. Node 35 cannot answer the
questions due to lack of knowledge about users pervious activities and it is marked as clone
nodes.
4.2.The role of

to constructing communities
In this section, the is set by some other values as represent in Table 4. For example when =1,
the number of augmented edges will be |E|. The clustered graphs with changing areshown in
Figure11. If is increased and the form of clustering does not change, it means that the default
value for was correct and most similar users are in each community.
Table 5. Different values of

K
0.68 34
0.78 39
0.88 44
1 50

88
Figure 11. The different graphs with different
For indicating the role of to construct communities with similar members, a similarity rate in
cluster parameter is defined as follows:
45467849: 879! 4; +6=9!8
5?!8 @A 7B5!;9!C !CB!= 4; +6=9!8
5?!8 @A !CB!= 4; +6=9!8
Figure 12. Similarity rate in community
The similar rate in clustersfor C1, C2 and C3 in Figure 10, is indicated in diagram of Figure 12.
According to diagram, through increasing the value of the most accurate clusters are obtained in
the light of similar members.
0
0.2
0.4
0.6
0.8
1
0.68 0.78 0.88 1
similarity
rate
in
cluster

C1
C2
C3

An IAC Approach for Detecting Profile Cloning in Online Social Networks

More Related Content

What's hot (19)

Similar to An IAC Approach for Detecting Profile Cloning in Online Social Networks (20)

Recently uploaded (20)

An IAC Approach for Detecting Profile Cloning in Online Social Networks