Secure Computation Laboratory

Constructing Stealth Attacks on Constrained Control Systems

Most cyber-physical systems are physically or economically connected and constrained. Model predictive control provided a systematic means of handling all forms of connections and constraints leading to tremendous impact on industrial control practice. MPC transcribes the control objective into a cost function and solves the constrained optimization online, with provable boundedness and stability.

Consequently, as more and more critical infrastructures that use some form of MPC like constrained controller such as aerospace, power and agricultural systems are being embedded with sensing and control and linked to the internet, the resulting security vulnerability can be exploited to inflict systematic damage to the connected physical systems. I shall construct false-data injection attacks that are able to confiscate control and remain stealthy to set-membership based anomaly detectors by masquerading as disturbances. These methods are used to demonstrate the vulnerability of a flight control system.