Resilient Collaborative Target Tracking Using BAG-aware Distributed Kalman Filtering Under Stealthy FDI Attacks

This paper presents a distributed framework for multi-agent 3D target tracking in the presence of False Data Injection (FDI) attacks. The proposed approach integrates Bayesian Attack Graph (BAG) modeling with Extended Kalman Consensus Filtering (EKCF) to maintain tracking performance of maneuvering aerial targets under persistent cyber-threats. We model attack propagation through a directed BAG, capturing the cascading nature of attacks in networked systems. Posterior probabilities of node compromise are computed by applying Bayesian inference over the node state using detection observations. These compromise probabilities inform the distributed estimation process, where agents with high compromise likelihood have their measurements rejected while uncompromised agents share their target estimates with their followers. Each sensing agent employs local EKCF for nonlinear target state estimation and Linear Quadratic Regulator (LQR) control for pursuit. The framework handles targets following 3D “Dubins" like airplane kinematics while agents are modeled as 3D double integrators. Simulation results demonstrate the system’s ability to maintain tracking accuracy comparable to attack-free scenarios, with root mean square errors remaining near the measurement noise level despite ongoing FDI attacks. The results validate the effectiveness of BAG-driven trust-based information fusion for cyber-resilient multi-agent systems tracking maneuvering targets in three-dimensional space.