Peer profile based trust model for P2P systems using genetic algorithm

The open and anonymous nature of P2P allows peers to easily share their data and other resources among multiple peers, but the absence of a defensible border raise serious security concerns for the users. There is a lack of accountability for the content that is shared by peers and it is hard to distinguish malicious users from honest peers. Establishing Trust relationship between peers can serve as the metric to determine the veracity of the shared content and reliability of the peers. Most of the research work in this area is on Reputation based trust management where trust is determined on the basis of recommendation of other peers. Such recommendations are subjective and can be biased. A number of peers can also collude to provide false testimony for malicious peers. This paper proposes a novel Trust model that combines peer profiling with anomaly detection technique. Each peer can establish trust based on its own prior activities with other peers by comparing the current activity of a peer with its historical data and Genetic Algorithm (GA) has been employed to detect the anomalous behavior. Peer profile is updated dynamically with every transaction using GA operator’s crossover and mutation. This model has been tested using a file sharing application against common attacks and the results obtained are compared with statistical anomaly detection approach.