Preventing Sybil attacks in P2P file sharing networks based on the evolutionary game model

Abstract

In cooperative Peer-to-Peer (P2P) networks, a number of users, called Free-riders, try to receive service from others without cooperating with them. Some others, called Sybil nodes, break the rules of the system by colluding and showing fake identities. P2P networks are highly vulnerable to these attacks. In previous research, no method has been suggested to counter these two attacks simultaneously. In the proposed method, a new centrality relationship has been used in the incentive mechanism to deal with both problems at the same time. In this regard, the more varied the nodes receiving service from a peer are, the better the peer reputation will be. The results show that the longer the network life goes on, the more free-riders are detected, and the number of services delivered to the collusive nodes will also be reduced.

Introduction

Peer-to-Peer (P2P) networks are considered as an effective way to organize distributed systems, which allow a group of users to communicate with each other and share their resources. They are mainly built on a network like the internet. Deploying networks such as CAN, (Content Addressable Network), Chord, Pastry and Tapestry on a large scale has enabled millions of nodes (users) to be able to share their data [32].

P2P architecture was a revolution in sharing large files on the internet. This architecture provided the opportunity for each peer to contribute to both uploading and downloading files. This contribution is autonomous, and each user decides on which level to contribute. The success of a P2P file sharing system depends on the contribution of users [18].

The open and dynamic nature of P2P networks can be both useful and dangerous. Problems such as free-riders and malicious users can cause a lot of problems in the proper functioning of the system [8].

In this infrastructure, nodes provide or use resources. The nodes can request the service from others or provide a service for others. Each node acquires benefit by receiving the service and pays a price by providing it for other nodes. In this cooperative model, each node usually tries to receive the most service possible. Defectors, also known as free-riders, are only looking to download their own resources among shared ones. They also avoid offering services to other nodes [7], [37].

The spread of such a phenomenon can be destructive. It might also reduce the value of file sharing in the network and further turn it into a sick network. According to a study conducted in 2005, 85% of Gnutella network users are free riders, and only 1% of the users share files and resources spontaneously [37].

Unfortunately, the P2P system does not have a central controller capable of monitoring user performance. Therefore, detection and prevention of malicious behaviors in this environment has become a major challenge. How to manage nodes and encourage them to cooperate is a fundamental issue in this open system. Trust and reputation techniques are the key ways to create collaborating behavior in P2P systems. Many solutions have so far been proposed based on these techniques, each with their own trust estimation, reputation dissemination and response to non-cooperative behavior methods [27].

In decentralized, distributed and uncontrolled systems, a user can pose a problem for the system by gaining and controlling a large number of IDs [28], [38]. This type of attack is called a Sybil attack and is known as a major threat to P2P systems. In this attack, a user with several fake identities exists on the network. Douceur [15] has proven that it is impossible to completely eliminate Sybil nodes, thus great efforts were made to minimize the malicious effects of the Sybil attack.

Much research has been done to identify free riders in P2P file sharing networks. Some of the most important papers in this area have been discussed in Section 2. In all these researches, a method has been proposed to detect the presence of a free rider in the network. After detecting the free riders, providing service for them can be stopped, or the quality of the service can be reduced. This will create an incentive for network users to cooperate. On the other hand, other papers focused on the identification of Sybil attacks. Some of these papers have been reviewed in Section 2. These researches focus on identifying users who are trying to violate the rules governing the network.

Accordingly, the main questions in this research that we are looking for include the following:

• How can we design a mechanism that can detect free riders in as well as confront Sybil attacks?

• How can a robust incentive mechanism be designed to help users cooperate with the network?

• What parameters affect the network users’ incentive?

None of the previous methods simultaneously deal with free riders and Sybil nodes. In this paper, a reputation-based approach is proposed that can identify these two types of malicious behaviors simultaneously.

The rest of the paper is organized as follows: Section 2 contains previous relevant work. Section 3 contains the system model. Section 4 describes the structure of the proposed incentive mechanism and information on how to calculate the centrality. In Section 5, simulation results of this game are presented. Section 6 reports on the conclusion drawn from this research.