Abstract:While automated vehicles (AVs) are expected to revolutionize future transportation systems, emerging AV technologies open a door for malicious actors to compromise intelligent vehicles. As the first generation of AVs, adaptive cruise control (ACC) vehicles are vulnerable to cyberattacks. While recent effort has been made to understanding the impact of attacks on transportation systems, little work has been done to systematically model and characterize the malicious nature of candidate attacks. In this study, we develop a general framework for modeling and synthesizing two types of candidate attacks on ACC vehicles, namely direct attacks on vehicle control commands and false data injection attacks on sensor measurement, with explicit characterization of their adverse effects. Based on linear stability analysis of car-following dynamics, we derive a series of analytical conditions characterizing the malicious nature of potential attacks. This ensures a higher degree of realism in modeling attacks with adverse effects, as opposed to simply considering attacks as constants or random variables. Notably, the conditions derived provide an effective method for strategically synthesizing an array of candidate attacks on ACC vehicles. We conduct extensive simulation to examine the impacts of intelligently designed attacks on microscopic car-following dynamics and macroscopic traffic flow. Numerical results illustrate the mechanism of candidate attacks, offering useful insights into understanding the vulnerability of future transportation systems. The methodology developed allows for further study of the widespread impact of strategically designed attacks on traffic cybersecurity, and is expected to inspire the development of efficient attack detection techniques and advanced vehicle controls.

ML-driven Malware that Targets AV Safety

Data Poisoning Attacks in Internet-of-Vehicle Networks: Taxonomy, State-of-The-Art, and Future Directions.

Cyberattacks on Adaptive Cruise Control Vehicles: An Analytical Characterization

Physical Backdoor Attack can Jeopardize Driving with Vision-Large-Language Models

Partial-Information, Longitudinal Cyber Attacks on LiDAR in Autonomous Vehicles

Detecting stealthy cyberattacks on adaptive cruise control vehicles: A machine learning approach

Stop-and-Go: Exploring Backdoor Attacks on Deep Reinforcement Learning-based Traffic Congestion Control Systems

Strategic Safety-Critical Attacks Against an Advanced Driver Assistance System

An Analytical Framework for Modeling and Synthesizing Malicious Attacks on ACC Vehicles

Deep learning adversarial attacks and defenses in autonomous vehicles: a systematic literature review from a safety perspective

Adversarial Attacks Against Deep Reinforcement Learning Framework in Internet of Vehicles

Impact of cyberattacks on safety and stability of connected and automated vehicle platoons under lane changes

BRAUM: Analyzing and Protecting Autonomous Machine Software Stack

Who is in Control? Practical Physical Layer Attack and Defense for mmWave based Sensing in Autonomous Vehicles

The Impact of Cybersecurity Attacks on Human Trust in Autonomous Vehicle Operations

Discovery of AI/ML Supply Chain Vulnerabilities within Automotive Cyber-Physical Systems

Work-in-Progress: Crash Course: Can (Under Attack) Autonomous Driving Beat Human Drivers?

Collective behavior and emergent risks in a model of human- and autonomously-driven vehicles

Adversarial Sensor Attack on LiDAR-based Perception in Autonomous Driving

Physical Backdoor Trigger Activation of Autonomous Vehicle using Reachability Analysis