Physical layer attack identification and localization in cyber-physical grid: An ensemble deep learning based approach

The massive integration of low-cost communication networks and Internet of Things (IoT) in today's cyber-physical grids has been accompanied by significant concerns regarding potential security threats. Specifically, wireless communication technology introduces additional vulnerability in terms of network security. In addition to cyber-security issues that have been investigated extensively, we must consider physical layer security. As such, considerable efforts have been employed toward developing a solution to address cyber-security issues. However, there are limited efforts on developing intrusion detection systems for physical layer security. In this paper, we propose an intelligent attack detection and identification model capable of classifying the attack type in the physical layer based on an ensemble of machine learning methods. Furthermore, the proposed model localizes the attack or fault to specific features or measurements in the system to assist cyber-security professionals in mitigating the effect of the attack in communication networks. The proposed model is evaluated on a smart grids dataset simulated by the Oak Ridge National Laboratories and is compared with traditional machine learning classifiers. The localization of attacks and faults is tested by splitting the data and measuring the correlation of the localization metrics produced by the proposed model. The results demonstrate the effectiveness of the proposed method at classifying and localizing attacks compared to peer approaches. (C) 2021 Elsevier B.V. All rights reserved.