COMBINING DEPENDENT INFORMATION BY EVIDENTIAL NETWORK REASONING FOR SITUATION AWARENESS OF CYBER-PHYSICAL SYSTEMS

Many cyber-physical systems, such as smart grid or SCADA, contain multiple sensor clusters. Internet connected sensor networks such as these inherit the high dependability requirements of the underlying physical process, whilst still being vulnerable to security threats from the cyber domain. Control security systems are used to monitor the communication between a central controller and a single sensor in order to protect the system’s normal operations from cyber attacks. The status of each physical component (hosts) in a system defines a situation that the system may be in. Cyber-security sensors such as malware detection software and host-based intrusion detection systems are used to monitor the status of a host.

This project continues our existing work on evidential network modelling of cyber-physical system state inference. Sensors of different types provide information complementary to each other and yet with different reliabilities. This introduces uncertainty to the data. Also, the dynamic capabilities of different sensors for monitoring system states bring uncertainty to the sensor-state implication relationships. Evidential networks allow intuitive reasoning about uncertain information based on the Dempster-Shafer theory of evidence. However, several issues remain to be investigated further. This PhD project will focus on investigating and developing a novel mechanism for measuring the dependency of sensor evidence in sensor-state relation implication rules and combining them with a variety of degrees of dependency. We aim to apply this to autonomous automotive systems (driverless cars).

Contact Details

Dr. Paul Miller
Telephone: +44 (0)28 9097 4637
email: p.miller@qub.ac.uk