A Novel Deep Learning Based Intrusion Detection System for Smart Meter Communication Network

Smart meter requires the bidirectional communication network for the transmission of real time power usage details to control centre. Cyber security is a major challenge of smart meter communication network. The direct communication of customer with end home devices increases the probability of attacks on the network. Intrusion detection system (IDS) using machine learning algorithms is one of the widely used attacks detecting mechanism. The large number of multivariate data generated from the smart meters affects the attack detection of IDS. In this work, a novel attack detection system using deep learning algorithms are proposed for accurately detects the attacks by analyzing the smart meter traffic. The proposed system has multiple multi layer deep algorithms, arranged in hierarchical order to detect the attacks accurately. The performance of the proposed system is analyzed by comparing with simple multi layer deep learning algorithm and hierarchical SVM algorithms with feature selection methods using the standard CICIDS 2017 dataset.