An Anomaly Behavior Analysis Framework for Securing Autonomous Vehicle Perception

TL;DR

The paper tackles the security of autonomous vehicle perception by introducing an Anomaly Behavior Analysis framework that learns normal perception behavior from a physics-based vehicle model and temporal features, combining model-based reasoning with supervised learning to detect attacks. It validates the approach through a depth-camera blinding attack on a QCar testbed and introduces the AVP-Dataset for public use. Key contributions include the integration of a Dynamic Bicycle Model with a data-driven detector and a publicly available dataset to benchmark intrusion-detection techniques, demonstrated by state-estimation accuracy, superior classifier performance (Random Forest), and effective threshold tuning that minimizes false alarms. The work offers a practical path toward perception-security in AVs by enabling continuous monitoring of perception behavior and providing a resource for researchers to evaluate detection strategies in real-world attack scenarios.

Abstract

As a rapidly growing cyber-physical platform, Autonomous Vehicles (AVs) are encountering more security challenges as their capabilities continue to expand. In recent years, adversaries are actively targeting the perception sensors of autonomous vehicles with sophisticated attacks that are not easily detected by the vehicles' control systems. This work proposes an Anomaly Behavior Analysis approach to detect a perception sensor attack against an autonomous vehicle. The framework relies on temporal features extracted from a physics-based autonomous vehicle behavior model to capture the normal behavior of vehicular perception in autonomous driving. By employing a combination of model-based techniques and machine learning algorithms, the proposed framework distinguishes between normal and abnormal vehicular perception behavior. To demonstrate the application of the framework in practice, we performed a depth camera attack experiment on an autonomous vehicle testbed and generated an extensive dataset. We validated the effectiveness of the proposed framework using this real-world data and released the dataset for public access. To our knowledge, this dataset is the first of its kind and will serve as a valuable resource for the research community in evaluating their intrusion detection techniques effectively.

Figures (5)

• Figure 1: Anomaly Behavior Analysis Methodology
• Figure 2: Dynamic Bicycle Model of a vehicle in a 2-dimensional inertial frame
• Figure 3: Architecture of the experimental setup
• Figure 4: Estimated dynamics vs actual dynamics of QCar. Plot limited to instances 5000-5200 for improved visualization
• Figure 5: Comparison of probability score distribution for normal and abnormal data. In 1 fold of test set: Normal data: 3893 instances, abnormal data: 13,902 instances