[Paper Review] Counteracting Byzantine Adversaries with Network Coding: An Overhead Analysis

Minji Kim, Muriel Médard+1 more|ArXiv.org|Jun 27, 2008

Cooperative Communication and Network Coding9 references4 citations

TL;DR

This paper analyzes the transmission overhead of three Byzantine adversary detection schemes in network-coded systems: end-to-end error correction, packet-based detection, and generation-based detection. It shows that when attack probability is high, packet-based detection minimizes bandwidth overhead by discarding corrupted packets early; when attack probability is low, generation-based detection is more efficient due to lower per-packet hash costs. The key contribution is a quantitative tradeoff analysis between generation size, detection overhead, and attack probability, demonstrating that detection schemes can outperform naive error correction under high adversarial conditions.

ABSTRACT

Network coding increases throughput and is robust against failures and erasures. However, since it allows mixing of information within the network, a single corrupted packet generated by a Byzantine attacker can easily contaminate the information to multiple destinations. In this paper, we study the transmission overhead associated with three different schemes for detecting Byzantine adversaries at a node using network coding: end-to-end error correction, packet-based Byzantine detection scheme, and generation-based Byzantine detection scheme. In end-to-end error correction, it is known that we can correct up to the min-cut between the source and destinations. However, if we use Byzantine detection schemes, we can detect polluted data, drop them, and therefore, only transmit valid data. For the dropped data, the destinations perform erasure correction, which is computationally lighter than error correction. We show that, with enough attackers present in the network, Byzantine detection schemes may improve the throughput of the network since we choose to forward only reliable information. When the probability of attack is high, a packet-based detection scheme is the most bandwidth efficient; however, when the probability of attack is low, the overhead involved with signing each packet becomes costly, and the generation-based scheme may be preferred. Finally, we characterize the tradeoff between generation size and overhead of detection in bits as the probability of attack increases in the network.

Motivation & Objective

• To evaluate the transmission overhead of three Byzantine detection schemes in network-coded systems: end-to-end error correction, packet-based detection, and generation-based detection.
• To determine under what conditions Byzantine detection schemes outperform traditional error correction by reducing bandwidth waste from corrupted data.
• To characterize the tradeoff between generation size and detection overhead as a function of attack probability.
• To assess whether the computational and communication cost of detection is justified by improved network throughput in adversarial environments.

Proposed method

• Models a multicast network using random block linear network coding with generation size G and packet size n.
• Analyzes three schemes: (1) end-to-end error correction, (2) packet-based detection using per-packet hashes of size hp, and (3) generation-based detection using per-generation hashes of size hg.
• Derives expected overhead as the ratio of corrupted or dropped bits to total received bits, considering attack probability p and hash costs.
• Uses probabilistic analysis to compute expected number of corrupted packets per generation and expected number of dropped packets under each scheme.
• Compares schemes by evaluating the ratio of expected overhead to total bits received, with sensitivity to p, G, and hash size.
• Employs asymptotic analysis to show that for p > 0.5, expected throughput approaches zero, making overhead negligible.

Experimental results

Research questions

• RQ1 Under what conditions does Byzantine detection reduce network overhead compared to end-to-end error correction?
• RQ2 How does the choice between packet-based and generation-based detection impact bandwidth efficiency as a function of attack probability?
• RQ3 What is the optimal generation size that minimizes detection overhead for a given attack probability?
• RQ4 At what attack probability does the cost of per-packet hashing in packet-based detection become justified compared to generation-based detection?
• RQ5 Can Byzantine detection schemes improve network throughput when the number of attackers is high, despite added detection overhead?

Key findings

• When the attack probability p exceeds approximately 0.03, the overhead of packet-based detection becomes less than that of end-to-end error correction, making it more bandwidth-efficient.
• For low attack probabilities (p < 0.03), the cost of per-packet hashing in packet-based detection exceeds the benefit, making end-to-end error correction more efficient.
• The generation-based detection scheme performs best at low attack probabilities because it spreads the hash cost over G packets, reducing per-packet overhead.
• At moderate attack probabilities (p ≈ 0.2), the generation-based scheme suffers peak overhead due to discarding entire generations even with few corrupted packets.
• For p > 0.5, the expected throughput drops to near zero, so overhead is negligible as no data is transmitted.
• The packet-based scheme is most bandwidth-efficient when attack probability is high, as the cost of hashing is outweighed by avoiding transmission of corrupted packets.