Secure Distributed State Estimation for Large-Scale Systems
Primary Investigator:
Shreyas Sundaram
Aritra Mitra and Shreyas Sundaram
Abstract
Real-time monitoring of large-scale complex systems such as power grids, gas and steam turbines, and air-traffic management systems is typically achieved by collecting and analyzing real-time plant data from sensors distributed over a communication network. Failures in such sensors can have catastrophic effects, especially in safety-critical systems. Consequently, securing these sensors against adversarial attacks (in the form of false data injections, denial of service attacks etc.) is a challenging and important problem. To this end, we develop a secure, lightweight, distributed algorithm for asymptotically reconstructing the state dynamics of an LTI system at each sensor node of a given network, in the face of worst-case adversarial attacks on some of the sensors. We also provide provable guarantees regarding the success of our proposed algorithm and characterize network topologies that facilitate our strategy. Our state estimation scheme provides fundamental insights into the connection between the plant dynamics, the sensor measurement structures and the underlying communication network topologies, and is expected to significantly enhance real-time data analytics.