FDIA generation design aginst L2 estimators (including least-square observer, Luenberger observer, Kalman filter) have been well studied, and motivates model-based and learning-based approaches for attack generation problem.
However, we recently observer the ineffectiveness of those FDIA designs against moving-horizon estimators (MHE). Thus, we are sharing a heuristic moving-hroizon FDIA design framework and algorithm in our paper:
Moving-horizon False Data Injection Attack Design against Cyber-Physical Systems (under review).
In the scenario of MHE, we will call "static state estimators" as those estiamtors which are designed with window's size equal to 1 (including those L2 estimators we mention at the beginning).
We share the simulation of MH-FDIA design here:
In this repo, I am presenting two demos to validate the proposed MH-FDIA:
- linear control system of IEEE 14-bus system (see branch demo_14_bus_system)
- nonlinear path-tracking control system of differential--driven mobile wheeled robot (see branch demo_suto_vehicle)