Cyber-physical system security

Due to an increasing reliance on IoT solutions, unsuspecting cyber-physical systems such as HVAC systems in buildings, are becoming the target of cyberattacks all over the world. Many of the existing cybersecurity tools and approaches focus on IT systems (e.g., network traffic and denial of service) and the effectiveness of such tools to secure cyber-physical systems is not clear.  Given that cyber-physical systems operate under physical laws specific to each domain, it would be possible to detect cyberattacks by incorporating domain models within data-driven AI approaches. The challenge then is to distinguish cyberattacks from equipment faults and degradation of a piece of equipment. This research project will focus on development of novel cyber-physical attack detection algorithms through integration of domain models with data driven AI/ML approaches. Specific focus will be on HVAC systems in buildings, which are known to be especially vulnerable to cyber-attacks, and testing the algorithms to be developed in real-life settings on a testbed at Carnegie Mellon campus.