Improving trust valuation for file sharing in P2P networks

Many peer-to-peer networks are used to share different types of files. These files could be damaged, contain false content or even include malware. To be able to choose the right peers, the requesting peer needs to assess the trustworthiness of the requested peer. Embedded trust algorithms enable the requesting peers to perform this assessment without the continuous involvement of human users. Requesting peers can then autonomously calculate the trustworthiness of the other peers before deciding to download. In this paper we show that current trust algorithms suffer from different shortcomings. Based on these results, we designed a new hybrid trust algorithm. We offer a step-by-step description of its working and present and analyse the results of a series of simulation tests.

I. Introduction

Peer to peer (p2p) networks are widely used nowadays with as main application file sharing. Examples include file sharing systems like BitTorrent, Napster and Gnutella. When a peer joins a peer-to-peer (p2p) network, it becomes part of it. If the peer intends to share files, it will have to connect and commit transactions (i.e., sharing (parts of) files) with other peers in the network. Often, the peers in the network are strangers to each other, in the sense that most of them have never been connected before [7]. Thus, the problem is to know how to choose the right peers, in order to avoid (file sharing with) the malicious ones. One way to deal with this security threat is to accept only files from trustworthy peers. We here focus on autonomous trust valuation of the file-sharing peers. Such a trust algorithm calculates the trust of a trustor peer on the basis of certain peer properties. The outcome of the computation process can be expressed as a number and labelled as a trust value describing the trustworthiness of the peer.