Enhancing cyber-physical security through data patterns

In this position paper, we propose a data-driven approach for security management in a network that interacts or receives inputs from physical systems - including human behavior. Our goal is to leverage the unique features of cyber-physical systems. In particular we propose: (1) the use of historical data from physical systems and human behaviors to enable anomaly detection, (2) the use of contextual data from multiple and diverse sensor readings to obtain a higher-level collective vision of the network for better event correlation and decision analysis, and (3) the use of physical sensor data and human behavior to enable fine-grained, dynamic access control and implicit authentication. We outline use cases describing how our ideas can be applied in the Home Area Network (HAN).