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:

@article{ZHENG2023105552,
title = {Moving-horizon false data injection attack design against cyber–physical systems},
journal = {Control Engineering Practice},
volume = {136},
pages = {105552},
year = {2023},
issn = {0967-0661},
author = {Yu Zheng and Sridhar Babu Mudhangulla and Olugbenga Moses Anubi},
doi = {https://doi.org/10.1016/j.conengprac.2023.105552}
}

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:

1. linear control system of IEEE 14-bus system (see branch demo_14_bus_system)

2. nonlinear path-tracking control system of differential--driven mobile wheeled robot (see branch demo_auto_vehicle, experiment_auto_vehicle)