A Physics-Informed Context-Aware Approach for Anomaly Detection in Tele-driving Operations Under False Data Injection Attacks

TL;DR

The paper addresses the security of Tele-operated Driving (ToD) by modeling False Data Injection (FDI) attacks on steering commands and proposing a Physics-informed Context-Aware Anomaly Detection System (PCADS). It introduces a dual-stage detection framework: first leveraging driving contexts to validate intended maneuvers, then applying a physics-informed LSTM to verify physical signatures against maneuvers using vehicle dynamic parameters. The work presents a threat analysis, an FDI attack formulation, and experimental validation across route planning, intersection maneuvers, and drivetrain variants, showing that physics-informed detection significantly outperforms input-based approaches. The findings suggest a robust defense-in-depth approach for ToD security, with practical implications for real-time monitoring and logging under UN R155 regulations, and point to future work on richer attack datasets and cross-domain comparisons.

Abstract

Tele-operated driving (ToD) systems are special types of cyber-physical systems (CPSs) where the operator remotely controls the steering, acceleration, and braking actions of the vehicle. Malicious actors may inject false data in communication channels to manipulate the tele-operators driving commands to cause harm. Hence, protection of this communication is necessary for the safe operation of the target vehicle. However, according to the National Institute of Standards and Technology (NIST) cybersecurity framework, protection merely is not enough and the detection of an attack is necessary. Moreover, UN R155 mandates that security incidents across vehicle fleets be detected and logged. Thus, cyber-physical threats of ToD are modeled with an attack-centric approach in this paper. Then, an attack model with false data injection (FDI) on steering control commands is created from real vehicle data. The risk of this attack model is assessed for a last-mile delivery (LMD) application. Finally, a physics-informed context-aware anomaly detection system (PCADS) is proposed to detect such false injection attacks, and preliminary experimental results are presented to validate the model.

Figures (13)

• Figure 1: A workflow of the paper's contributions.
• Figure 2: An attack tree for a ToD event.
• Figure 3: A tele-operated vehicle attack likelihood vs impact.
• Figure 4: An FDI attack on communication between remote operators and the vehicle.
• Figure 5: A traffic light intersection attack scenario.