Reputation and Audit Bit Based Distributed Detection in the Presence of Byzantine

TL;DR

This paper introduces two reputation-based clustering algorithms, RAC and RACA, that effectively detect Byzantine nodes in distributed networks without prior knowledge of attack strategies, maintaining high performance even under dynamic conditions.

Contribution

The paper proposes novel reputation and audit-based clustering algorithms that improve Byzantine attack detection in distributed networks without requiring prior attack information.

Findings

Both algorithms outperform existing methods in detection accuracy.

RACA remains effective even when Byzantine nodes exceed half the network.

Algorithms maintain robustness in dynamic, changing attack scenarios.

Abstract

In this paper, two reputation based algorithms called Reputation and audit based clustering (RAC) algorithm and Reputation and audit based clustering with auxiliary anchor node (RACA) algorithm are proposed to defend against Byzantine attacks in distributed detection networks when the fusion center (FC) has no prior knowledge of the attacking strategy of Byzantine nodes. By updating the reputation index of the sensors in cluster-based networks, the system can accurately identify Byzantine nodes. The simulation results show that both proposed algorithms have superior detection performance compared with other algorithms. The proposed RACA algorithm works well even when the number of Byzantine nodes exceeds half of the total number of sensors in the network. Furthermore, the robustness of our proposed algorithms is evaluated in a dynamically changing scenario, where the attacking parameters change over time. We show that our algorithms can still achieve superior detection performance.