Defending Against sybil-attacks in Peer-to-Peer Networks

Xiang Xu, Huijuan Lu, Lianna ChenInternational Journal of Security and Its Applications 2014 · DOI

Abstract

Peer-to-Peer (P2P) networks have become a popular way for users to share files over the Internet. However, there has been a spurt of works showing that the existence of Sybil attacks is a serious threat to P2P networks, where one or more attackers can forge a large number of fictitious identities. Implementing correct protocols to address Sybil attacks is the key to improving performance. In this paper, we present a novel system to defend against Sybil attacks. Our direct and indirect transaction protocols limit the number of service units that a node can obtain. Furthermore, we design a dynamic reputation ranking algorithm for the indirect transaction protocol. Combining these two, a node with a high priority has more probability of obtaining service. Our system does not try to prevent users from creating multiple identities, but they cannot gain extra profit from doing so. It achieves a provable performance and overcomes the limitation of current social network-based defenses. Simulation results show that our system achieves a provable bound in terms of the number of service units obtained by Sybil nodes while not sacrificing the performance of the file sharing application.

Introduction

P2P networks, such as those used for file sharing, rely heavily on the cooperation of their participants. However, they are frequently undermined by two major issues: free riding (users who download but refuse to upload) and Sybil attacks. In a Sybil attack, a single adversary creates a massive number of fake identities to manipulate reputation systems or bypass resource limits.

Traditional defenses often rely on a central authority, which creates a single point of failure, or social network analysis, which can be inaccurate if honest communities are sparsely connected. This paper proposes a decentralized solution that works on both structured and unstructured networks without needing to "detect" communities, instead focusing on making attacks economically unprofitable.

Methodology

The authors model the P2P network as a directed graph where edge weights represent the "net contribution" between two nodes. If node $A$ has uploaded 10 units to node $B$ and downloaded 2, the edge weight $R_{A,B}$ is 8.

Direct Transaction Protocols

The system uses two primary modes for direct interactions:

  • Supply Mode: Used for "bootstrapping" new connections. It has low priority.
  • Return Mode: High priority. If node $A$ previously helped node $B$, node $B$ is obligated to return the favor up to the amount previously received.
Figure 1: Illustration of the Direct Transaction Protocol. (a) shows an initial credit of 5. (b) shows the state after a transaction of 8 units, resulting in a weight of -3.

Indirect Transaction Protocol (Remote Supply)

When two nodes are not direct neighbors, they find a path using a shortest-path algorithm. A "free parameter" ($b$) is introduced to tune the system. When a transaction happens across a path, the weights of all edges on that path are reduced by $t/b$ (where $t$ is the service units). This effectively "consumes" the trust along the path.

Figure 6: Indirect transaction protocol examples. (a) shows the initial path weights. (b) illustrates weight reductions for different values of service units (t) and the tuning parameter (b).

Dynamic Reputation Ranking Algorithm (DRRA)

When a node receives multiple requests simultaneously, it uses DRRA to decide whom to serve. The algorithm calculates a rank based on the weights of the path from the requester. Nodes that have contributed more to the network (higher edge weights) receive a better rank and higher priority.

Experimental Evaluation

The system was evaluated using a custom event-driven simulator built in C++ based on the Gnutella protocol specification. The synthetic network consisted of 4,096 nodes with initial degrees set to $d=8$. Edge weights were randomly initialized between -100 and 100 to simulate a pre-existing network state.

Key Results

  • Provable Security: The authors provide a mathematical bound showing that the service units a Sybil region can obtain is strictly limited by their contribution plus a factor of the number of attack edges multiplied by the maximum edge weight ($S_{NS} \le b \times S_{SN} + a \times T$).
  • Efficiency: In simulations with a path length of 6 and the parameter $b=2$, over 40% of service requests were successful, suggesting the system maintains good performance for honest users while under attack.
  • DRRA Effectiveness: Under high-load scenarios, the DRRA algorithm significantly outperformed basic protocols, ensuring that honest, high-contributing nodes were prioritized over Sybil nodes.

Conclusion

The core innovation of this system is the use of economic disincentives over detection. By strictly limiting a node's ability to receive service based on its historical contribution, a Sybil attacker gains no advantage from creating multiple identities. The security hinges on the "attack edges"—the few links an attacker manages to establish with honest nodes. Because every transaction reduces the weight of these edges, the attacker eventually runs out of credit unless they actually provide service to the honest network. Future work will focus on the trade-off between security and efficiency inherent in the $b$ parameter and reducing path discovery overhead in dynamic networks.

Summary updated

Abstract

Peer-to-Peer networks have become a popular way for users to share files over the Internet.However, there has been a spurt of works showing that the existence of sybil attacks is a serious threat to Peer-to-Peer networks, where one or more attackers can forge a large number of fictitious identities. Implementing correct protocols to address sybil attacks is the key to improving the performance.In this paper, we present a novel systemto defend against Sybil attacks.Our direct and indirect transaction protocols limit the number of service units that a node can obtain.Furthermore, we design a dynamic reputation ranking algorithm for the indirect transaction protocol.Combining these two, a node with a high priority has more probability of obtaining service.Our system does not try to prevent users from creating multiple identities, but they cannot gain extra profit from doing so.It achieves a provable performance and overcomes the limitation of current social network-based defenses.Simulation results show that our system achieves a provable bound in terms of the number of service units obtained by sybil nodes while not sacrificing the performance of the file sharing application.
BibTeX
@article{xu2014defending,
  author = {Xiang Xu and Huijuan Lu and Lianna Chen},
  title = {{Defending Against sybil-attacks in Peer-to-Peer Networks}},
  journal = {International Journal of Security and Its Applications},
  year = {2014},
  doi = {10.14257/ijsia.2014.8.4.30},
}
Figure