O impacto de calotes e múltiplas personalidades no BitTorrent.

Abstract

BitTorrent, one of the most popular content distribution protocols nowadays, has an identification generation scheme that is completely autonomous. Furthermore, BitTorrent uses a random mechanism to discover new peers. This leaves the system vulnerable to a sybil attack, by which an entity associates multiple identifications to itself in an attempt to fool the other peers that execute the agreed protocol and increase its utility. In this work we evaluate the impact of such an attack. We present an analytical model and validate it through simulations. Our initial analysis uses two metrics. The first one is the number of different identifications that an attacker must have in order to experience download times (utility) equal or smaller to that experienced by the other peers that collaborate resources to the system. The other metric assesses how faster an attacker downloads a file, compared to collaborators, as we increase the number of identifications that the attacker has. Our results show that the number of different identities required is relatively small. In addition, our analysis observed the dynamic of the interactions among the peers aiming at studying the behavior of attacks in typical swarms of a community that shares files for free distribution. Our results show that, in general, the attack is effective when considering the majority of the studied scenarios. In other few scenarios, it depends on the moment of the swarm’s life and on the number of identities.