Arindam Paul

Detecting Sybil in P2P networks using Psychometric Methods

Peer to peer networks guarantee complete user anonymity to the clients. However, modern P2P networks suffer from Sybil attacks, which forge multiple identities to influence the global decisions in the network. In a Sybil attack, an entity in a peer-to-peer network masquerades itself as multiple simultaneous identities in the network. Thus, a single user can exert a significant effect on the decisions or working of the entire network if the multiple identities created by the user form a significant fraction of the peer-to-peer network. It is difficult to resolve Sybil attack at identity level because they need not be doing anything wrong.

Psychometric tests help in characterizing a personality. Our work attempts to identify the Sybil groups using psychometric ratings. The psychometric ratings are based on the answers to questions based on personal psychological nature. Since a single malicious user is creating multiple identities, he will be required to answer multiple questionnaires on behalf of the identities he has created, in contrast to a single honest user. Our solution is based on analyzing the questionnaires received from all the identities and clustering them based on the common psychological characteristics. The clusters are suspected to be originating from the same malicious user. The suspicion is verified by use of CAPTCHAs.