Many critical infrastructure systems are operated using networked feedback control. These systems crucially use wireless networks to transmit sensor and actuation signals. Unfortunately, wireless technology (sensors, actuators and communications) is unreliable and increasingly vulnerable to cyberattacks. This causes performance degradation, loss of stability, system failure and, at worst, leads to deaths and disasters. Therefore, mitigating the effects of attack algorithms on Cyberphysical Systems (CPSs) is of utmost importance.
A distinguishing aspect, when compared to attacks on classical information systems, is that in CPSs attacks can destabilise the system leading to shutdowns or severe accidents. Crucially, attackers may have access to a CPS system model and control objectives. This knowledge can be used to carry out sophisticated attacks and requires equally (or more!) sophisticated mitigation mechanisms at a systems and control level.
This PhD project investigates how the resilience of CPSs against malicious attacks can be improved through a combination of control system design, and machine learning. This project expects to generate new knowledge on how attack patterns in CPSs can be detected (e.g., anomalous inputs), how the policy of the attacker can be learned, and how this knowledge can be utilised to mitigate attacks through control or reconfiguration of the CPS. Furthermore, the project develops a framework on how CPS can be designed from the ground up to make attacks easier to detect and mitigate.
This PhD project funded through our School’s Major Research Initiative and also involves A/Prof Niko Suenderhauf.
Two related publications: