Attack Identification for Nonlinear Systems Based on Sparse Optimization

Adversarial attacks on controllers of dynamic systems have become a serious threat to many real-world systems, making methods for fast identification an indispensable part autonomous systems. With the increasing use model-based controllers, it is valid exploit model knowledge also attack as long privacy individual components maintained. A scalable, method reveal generic was introduced in our previous work and further investigated here. It designed coupled with nonlinear dynamics monitors certain coupling states. Based exchange local sensitivity information, approximates propagation through network solves sparse optimization problem identify attack. We provide thorough derivation approach analyze involved approximation errors prove rigorous guarantees successful identification. In extensive numerical case study IEEE 30 bus power system, we that not only apply nonlinear, practically relevant example, but identifies attacked buses very high success rates.