Real-time Cyber-Physical Security Solution Leveraging an Integrated Learning-Based Approach

Cyber-Physical Systems (CPS) has emerged as a paradigm that connects cyber and physical worlds, which provides unprecedented opportunities to realize intelligent applications such smart home, cities, manufacturing. However, CPS faces great number of information security challenges (e.g., attacks) due the integration well human behaviors interactions. Therefore, accurate real-time attack detection identification are essential ensure reliability CPS. In this paper, we propose novel integrated learning method accurately detects an system then identifies type in real time. Specifically, consider One-Class Support Vector Machine (OCSVM) model only relies on data from normal state for training achieve effective (i.e., or under-attack). If is detected be under-attack, develop Pairwise Self-supervised Long Short-Term Memory (PSLSTM) approach identify type, aims distinguish known types discover unknown new attacks. Lastly, experimental results show proposed achieves promising performances compared with conventional state-of-the-art learning-based benchmarks.