A CNN-based approach for anomaly detection in smart grid systems

Power grids are always a major center of attraction for attackers due to their large coverage area, complex technological infrastructure, remote access facilities, connectivity, and many more. The power grids that have bidirectional power flow, information interchange, and advanced communication technologies to improve the quality of power transmission and distribution are called smart grids. DoS attacks and false data injection attacks are the two major attacks on smart grids. This paper considers a false data injection attack considering noise reduction, state estimation, system parameters, and control system generation. Firstly, we measure the noise and mitigate it using the data matrix of the devices (sensors, generators, substations, and breaker systems). Secondly, state estimation uses the CNN to track the injected attack types. Finally, a control system is designed to provide anomaly patterns to the devices as input to prevent attacks in the future. The experimental results show that the proposed method achieves a higher anomaly detection rate with a lower error rate (approximately . 1.43%). The network performance parameters are providing significant results; throughput (12 Mbps), PDR (0.005%), execution time (1.43 s), and energy consumption are reduced by 12%–15% as compared to the existing research findings.