Compromised account detection in online social networks: A survey

Abstract

Summary Online Social Networks are an integral part of the modern society. Social networking sites are at the top rank in terms of usage out of the billions of websites hosted in the internet. The way in which criminal activities are interleaved to these networks are beyond our imagination. Securing these networks from cyber criminals and malicious attackers are a major concern nowadays. There are various security threats happening to the Online Social Network accounts, and account compromise is one among it. Account compromise can cause collateral damage to the user and to a mass community. Misusing the personal information of the users and exploiting their trust relationships within the network lead to serious problems. These compromised accounts are becoming a safe zone for criminals to launch attacks. The threats caused by these compromised accounts include spreading malware, spy advertisements, and espionage. In most cases, it becomes too late to detect that an account is compromised. Detection of these accounts is important in order to reduce the impact of damage. The detection of compromised Online Social Network account is the process of confirming whether the account used by an individual or by an organization has lost its control or not. This survey reviews the various methods that are found in the literature toward detecting compromised Online Social Network Accounts.

A multiview approach based on naming behavioral modeling for aligning chinese user accounts across multiple networks

“…The kth name in N (a) i M The numbers of user accounts in  (1) N The numbers of user accounts in  (2)…”

“…The set of alignment relationships between the user accounts in  (1) and  (2) A i,j The alignment relationships between the user accounts u (1) i and u (2) j TA B L E 1 Main notations Suppose there are two network  (1) and  (2) , the user account sets of  (1) and  (2) are  (1) = {u (1) 1 , u (1) 2 , … , u (1) M } and  (2) = {u (2) 1 , u (2) 2 , … , u (2) N }, respectively. Where u (1) i and u (2) j are the ith user account in  (1) and the jth user account in  (2) , respectively. M and N denote the numbers of user accounts in  (1) and  (2) .…”

“…Where u (1) i and u (2) j are the ith user account in  (1) and the jth user account in  (2) , respectively. M and N denote the numbers of user accounts in  (1) and  (2) . N (a) i denotes the set of names used by the u (a) i , and n (a) i,k is the kth name in N (a) i , where a ∈ {1, 2}.…”

“…And  = {A i,j } M×N denotes the set of alignment relationships between the user accounts in  (1) and  (2) , if u (1) i and u (2) j are (or are not) owned by the same user, then the value of A i,j is set as 1 (or 0), and we label it as "positive" ("negative"); and if we do not know whether u (1) i and u (2) j belong to the same user or not, the value of A i,j is unknown, and thus it is an unlabeled alignment relationship. The task of user alignment across  (1) and  (2) is to create a model and use it to predict the value of each unknown A i,j in . Specifically, the goal of our research is to propose a user alignment model, which can properly use the information extracted from the names of Chinese user accounts and achieve good performances on aligning these accounts.…”

“…The account name similarity is computed from the Levenshtein distance, which denotes the minimum number of single-letter edits (insertions, deletions, or substitutions) required to change one name into the other name. For two account names n (1) a,b and n (2) c,d , if their Levenshtein distance is LD(n (1) a,b , n (2) c,d ), then their similarity is computed as:…”