Abstract:This paper studies the secure and safe control problem of connected and automated vehicles (CAVs) with false data injection (FDI) attacks. A secure and safe controller with a novel surrounding vehicles’ state estimator and an attack detector is proposed. The state estimation for surrounding vehicles is collectively processed by combining the deep neural network-based predictions with model-based estimations. Additionally, a weight in the loss function is proposed for more accurate predictions of vehicles that are closer to the ego vehicle. For attack detection, a novel scheme that utilizes control outcomes to train detection actions is proposed. Different from the objectives of existing detectors, the reward for the proposed detector is designed to encourage the CAV to fully utilize the observations. A reward setting and the decision preference of the ego vehicle are theoretically analyzed. The effectiveness of the proposed algorithm is validated in an open simulation environment.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve This paper aims to address the issue of secure control of connected and autonomous vehicles (CAVs) under false data injection (FDI) attacks. Specifically, the paper focuses on the following aspects: 1. **Security**: How to ensure that autonomous vehicles can make correct decisions in the presence of FDI attacks, avoiding safety hazards caused by false data. 2. **Reliability**: How to design a reliable controller that can maintain system stability and performance even when an attack is detected. 3. **Prediction Accuracy**: How to combine deep neural networks and physical models to improve the prediction accuracy of the states of surrounding vehicles, especially when new targets appear without historical trajectories. 4. **Attack Detection**: How to design a new attack detection mechanism that can not only identify attacks but also adaptively utilize observation data based on control effects to reduce the impact of attacks on the system. ### Main Challenges 1. **New Target Prediction**: Existing prediction algorithms become ineffective when new targets appear without historical observation trajectories. 2. **Parameter Sharing**: To reduce the number of parameters, existing methods often use parameter sharing techniques, but this limits the predictor's ability to prioritize the ego vehicle. 3. **Attack Detection**: Most existing attack detection algorithms judge based on the relationship between observations and predictions. A common approach is to discard attacked data, which significantly affects control performance. ### Main Contributions 1. **Hybrid Prediction Algorithm**: A hybrid prediction algorithm combining neural networks and physical models is proposed. It collectively considers all prediction targets centered on the ego vehicle, intuitively solving the new target prediction problem and prioritizing prediction tasks. 2. **Novel Detection Method**: A new detection method is proposed to adaptively utilize observation data, minimizing the impact of FDI attacks on secure control. Additionally, a solution based on deep reinforcement learning (DRL) is proposed, and the corresponding neural network is designed. 3. **Perception Confusion Analysis**: The limitations brought by perception confusion are analyzed during detection, and two reward design schemes with different objectives are theoretically proposed. Through these contributions, the paper provides an effective method for achieving secure control of connected and autonomous vehicles under FDI attacks.

Secure and Safe Control of Connected and Automated Vehicles Against False Data Injection Attacks

SAID: State-aware Defense Against Injection Attacks on In-vehicle Network

Safeguard Vehicle Platooning Based on Resilient Control Against False Data Injection Attacks

Distributed Data-Driven Control for a Connected Autonomous Vehicle Platoon Subjected to False Data Injection Attacks

Event-based secure control for cyber-physical systems against false data injection attacks

Adaptive Resilient Control for Autonomous Vehicles Steering System Against False Data Injection Attacks

Event-Triggered Path Following of Autonomous Vehicles under False Data Injection Attacks: A Try-Once-Discard Protocol-Based Transmission Scheme

ETM-Based Fault-Tolerant and Intrusion-Tolerant Control for 2-D Planar Vehicular Platoon with Actual Traffic Scenarios

Secure Estimation and Attack Isolation for Connected and Automated Driving in the Presence of Malicious Vehicles

Real-time False Data Injection Attack Detection in Connected Vehicle Systems with PDE modeling

Detection‐based resilient control for cyber‐physical systems against two‐channel false data injection attacks

Optimal Controller Realizations against False Data Injections in Cooperative Driving

Adaptive Security Control against False Data Injection Attacks in Cyber-Physical Systems

An Autoencoder-Driven FDIA Detection Method for Virtual Coupling Systems of Heavy-Haul Trains

Secure Consensus of Multiagent Systems Against False Data Injection Attacks: A Distributed Model Predictive Control Approach

Secure Formation Control for Unmanned Surface Vehicles Against False Data Injection Attacks

Advanced Bad Data Injection Attack and Its Migration in Cyber-Physical Systems

Data-driven Control Against False Data Injection Attacks

A Physics-Based Context-Aware Approach for Anomaly Detection in Teleoperated Driving Operations Under False Data Injection Attacks

Attacks Detection and Security Control for Cyber-Physical Systems under False Data Injection Attacks