Neural Networks-Aided Insider Attack Detection for the Average Consensus Algorithm

Gangqiang Li,Shengli Zhang,Qiang Li,Sissi Xiaoxiao Wu

doi:10.1109/access.2020.2978458

Abstract

To support the big-data processing needs of large-scale deployments of smart devices, there is significant interest in moving from cloud-computing to multi-agent (fog-computing) models, given these algorithms scalability and self-healing properties with respect to nodes and link failures. However, these algorithms are often based on the average consensus primitive, which is, unfortunately, vulnerable to data injection attacks. Recognizing this challenge, this work proposes three novel methods for detecting and localizing adversarial nodes using neural network (NN) models. The methods proposed are based on fully distributed algorithms, wherein each node locally updates its local states by exchanging information with its neighbors without supervision. Compared to the state-of-the-art, the proposed approach leverages the automatic learning characteristics of NN to reduce the dependence on pre-designed models and human expertise in complex internal attack scenarios. Simulation results show that the NN-based methods can significantly improve the attacker detection and localization performance.

Highlights

Nowadays, distributive resource allocation is expanding its footprint and attracting worldwide research and development efforts [1]–[4]
We describe three types of novel features to be used for training the neural network (NN) — including the temporal difference NN (TDNN), spatial difference NN (SDNN), and a Fourier transform-based NN (FDNN)
DISCUSSION In this paper, we have proposed three new defense strategies for randomized gossip algorithm for average consensus based on NNs to detect and localize insider adversarial nodes

Summary

Introduction

Distributive resource allocation is expanding its footprint and attracting worldwide research and development efforts [1]–[4]. To achieve better responsiveness and scalability, multi-agent algorithms are considered a preferable alternative to cloudbased solutions for a range of applications such as real-time control, resource management, [13], [14], and resource allocation problems [15]–[17] When they aim at a coordinated response, the majority of these algorithms incorporate a peer-to-peer consensus primitive, that leads to a consistent globally optimum decision [18]–[20]. Each node independently runs a consensus protocol in an iterative manner, where at each iteration it exchanges local information with its neighbors and updates its local states Such algorithms are inherently robust against intermittent communication and provide a certain degree of privacy for the participating agents compared to cloud-based centralized solutions. Despite such appealing features, decentralized algorithms are inherently vulnerable to insider data injection attacks, since each node implements the computation and message passing protocol iterations without any supervision [21], [22]

Objectives

Results

Conclusion