Towards a Bio-inspired Real-Time Intrusion Detection in the Smart Grid

Industrial Control Systems in the Smart Grid network are increasingly utilizing the advantage offered by their interconnectedness through wireless sensors and smart devices. This leverage is not without its attendant setbacks as attackers are exploiting vulnerabilities in the Cyber-Physical Systems. Existing intrusion detection mechanisms that harness biological intelligence lack real-time actions. This paper developed a Danger Theory-based intrusion detection model for the Smart Grid by relying on the emerging dendritic cells algorithm, a flagship of the artificial immune system to optimize traditional (RF and XGBoost) and deep learning (RNN, DBN, GRU, DNN, and LSTM) algorithms. Datasets from the CyberGrid Testbed of Africa Centre of Excellence at Obafemi Awolowo University, Ile-Ife, Nigeria was used to demonstrate the model and evaluated using accuracy, latency, and false alarm rate. Experimental results show that RF, XGBoost, GRU, DBN, RNN, DNN, and LSTM returned the latency of 0.0000 ms, 0.000671 ms, 0.001342 ms, 0.0000 ms, 0.001133 ms, 0.001156 ms, and 0.001266 ms, respectively. These results demonstrate that RF, DBN, and RNN had outstanding latencies for layer1, layer2, and layer3 accordingly.