Detection of Fake Accounts in Social Network

Abstract

Social network is the basic platform of today‟s world to get connected with the people having same type of interest, perform similar activities or known to each other. Being on network, there is always a chance of getting data hacked by some fraudster. As the users are unaware of the attacker, they simply share their information over the network. The fraudster enters into the network after detecting the weaker node of the network that has maximum information on its node as compared to other weaker nodes and tries to create a link with that node. Once the fraudster enters into the network, he tries to build trust so that it can gain access to the information present over the network. Social network is not just Facebook or LinkedIn or twitter but it goes far beyond that, all kinds of transactional data where two objects are related to each other implicitly or explicitly falls under a network. We propose an algorithm which creates a label on every node based on their behaviour in the network whether it is genuine or fake account. In this we measure the strength of every node in the network and compared it with the trust score calculated for every node. Based on these two parameters we classify the nodes into their respective class. Once each node has labels according to their behaviour, it is easier for the user to detect whether to accept the friend request from such a node or ignore it.