Research Of Trust Models Based On D-S Evidence Theory For P2P Networks

P2P has developed quickly and been applied widely recently. In P2P network, peers are equal and can sign in and out dynamically and commute with each other anonymously. There are short of trusts between peers because of the equality, openness, anonymity of peers and having no reliable authority to manage and supervise these peers so that making against the development of P2P seriously. So how to establish a reliable and effective trust model to form trust relation which can resist aggressive behavior is an important issue in P2P network.These trust models make initial D-S evidence combination rule as their combination rule after analyzing traditional trust models especially about trust models based on evidence theory. It is well known that initial D-S evidence combination rule has many defects, repelling conflicted information, combining conflicted information badly and so on. So the trust model based on D-S evidence theory is improved. Firstly the rationality and reliability of the original D-S evidence combination method is enhanced by allotting the conflicted probability again and introducing weight coefficients. The improved D-S evidence combination rule is called as IDSECR. Then a trust model is established by this enhanced evidence combination method which is called DSETTM. In DSETTM, there are three kinds of trust value:local trust value, global trust value, comprehensive trust value. Local trust value is computed based on direct transaction history by basic probability assign function of IDSECR. Global trust value is computed by other peers'recommendation by enhanced evidence combination method. Comprehensive trust value is composed of local trust and global trust by belief factor. Transaction is successful or not that is decided completely by comprehensive trust value. Trust value is corrected by time ruler, time reduction function and belief factor when computing trust value. It is well known that trust value is only a simple number in traditional trust models and judging method is by comparing the numbers, but in this thesis, peer's trust value is a list for all kinds of statement. In proportion, the method of measuring trust value is not decided by comparing numbers, but by favorability rating of statements. This method is more flexible.DSETTM has been simulated and the results show that our trust model can evaluate peer's trust value and distinguish malicious peers effectively. DSETTM can get a higher transaction successful rate under attacking of different types of malicious peers.