Random linear network coding can be used in peer-to-peer networks to increase the efficiency of content distribution and distributed storage. However, these systems are particularly susceptible to Byzantine attacks. We quantify the impact of Byzantine attacks on the coded system by evaluating the probability that a receiver node fails to correctly recover a file. We show that even for a small probability of attack, the system fails with overwhelming probability. We then propose a novel signature scheme that allows packet-level Byzantine detection. This scheme allows one-hop containment of the contamination, and saves bandwidth by allowing nodes to detect and drop the contaminated packets. We compare the net cost of our signature scheme with various other Byzantine schemes, and show that when the probability of Byzantine attacks is high, our scheme is the most bandwidth efficient.
Deep Dive into On Counteracting Byzantine Attacks in Network Coded Peer-to-Peer Networks.
Random linear network coding can be used in peer-to-peer networks to increase the efficiency of content distribution and distributed storage. However, these systems are particularly susceptible to Byzantine attacks. We quantify the impact of Byzantine attacks on the coded system by evaluating the probability that a receiver node fails to correctly recover a file. We show that even for a small probability of attack, the system fails with overwhelming probability. We then propose a novel signature scheme that allows packet-level Byzantine detection. This scheme allows one-hop containment of the contamination, and saves bandwidth by allowing nodes to detect and drop the contaminated packets. We compare the net cost of our signature scheme with various other Byzantine schemes, and show that when the probability of Byzantine attacks is high, our scheme is the most bandwidth efficient.
Network coding [1], an alternative to the traditional forwarding paradigm, allows algebraic mixing of packets in a network. It maximizes throughput for multicast transmissions [2], [3], [4], as well as robustness against failures [5] and erasures [6]. Random linear network coding (RLNC), in which nodes independently take random linear combination of the packets, is sufficient for multicast networks [7], and is suitable for dynamic and unstable networks, such as peer-to-peer (P2P) networks [8], [9].
A P2P network is a cooperative network in which storage and bandwidth resources are shared in a distributed architecture. This is a cost-effective and scalable way to distribute content to a large number of receivers. One such architecture is the BitTorrent system [10], which splits large files into small blocks. After a node downloads a block, it acts as a source for that particular block. The main challenges in these systems are the scheduling and management of rare blocks.
As an alternative to current strategies for these challenges, [8], [9] propose the use of RLNC to increase the efficiency of content distribution in a P2P solution. These schemes are completely distributed and eliminate the need of a scheduler, since each node independently forwards a random linear combination. In addition, there is a high probability that each packet a node receives is linearly independent of the previous ones, and thus, the problem of redundancy caused by the flooding approaches in traditional P2P networks is reduced. RLNC based schemes significantly reduce the downloading time and improve the robustness of the system [8], [11].
Despite their desirable properties, network coded P2P systems are particularly susceptible to Byzantine attacks [12], [13], [14] -the injection of corrupted packets into the information flow.
Since network coding relies on mixing of packets, a single corrupted packet may easily corrupt the entire information flow [15], [16]. Furthermore, in P2P networks, there is typically no security control over the nodes that join the network and the packets that they redistribute. The topologies of the overlay graphs that arise from traditional P2P networks are often modeled as scale-free and small-world networks [17], [18], which are prone to the dissemination of epidemics, such as worms and viruses [19], [20]. Several authors address these problems in coded P2P networks.
We shall discuss these countermeasures in Section II. Most of these can be divided into two main categories: (i) end-to-end error correction and (ii) misbehavior detection.
Motivated by these observations, we address the issues of Byzantine adversaries in coded P2P networks. The main contributions of this paper are as follows:
• We propose a model for the evaluation of the impact of Byzantine attacks in coded P2P networks, and provide analytical results which show that, even for a small probability of attack, the information can become contaminated with overwhelming probability.
• We propose a new efficient, packet-based signature scheme, designed specifically for RLNC systems, to detect Byzantine attacks by checking the membership of a received packet in the valid vector space. This scheme allows an one-hop containment of the contamination.
• We analyze the overhead in terms of bandwidth associated with our signature scheme, and compare it to that of various Byzantine detection schemes. We also show that our scheme is the most bandwidth efficient if the probability of attack is high. This paper is organized as follows. Section II gives an overview of network coding in P2P networks and existing Byzantine detection schemes. In Section III, we analyze the impact of Byzantine attacks on the system. We propose our signature scheme in Section IV, and compare its overhead with other schemes in Section V. Finally, we conclude in Section VI.
References [6], [7] propose a random block linear network coding system -a simple, practical capacity-achieving code, in which every node independently constructs its linear code randomly. In such a system, a source generates information in batches of G packets (called a generation).
The source then multicasts them to its destination nodes using RLNC, where only the packets from the same generation are mixed. Note that RLNC is a distributed protocol, which requires no state information; thus, making it suitable for dynamic and unstable networks where state information may change rapidly or may be hard to obtain.
Several authors have evaluated the performance of network coding in P2P networks. Gkantsidis et al. [9] propose a scheme for content distribution of large files in which nodes make forwarding decisions solely based on local information. This scheme improves the expected file download time and the robustness of the system. Reference [8] compares the performance of network coding with traditional coding measures in a distributed storage setting with very limited storage space with the goal of mi
…(Full text truncated)…
This content is AI-processed based on ArXiv data.
