Abstract:In autonomous driving, the combination of AI and vehicular technology offers great potential. However, this amalgamation comes with vulnerabilities to adversarial attacks. This survey focuses on the intersection of Adversarial Machine Learning (AML) and autonomous systems, with a specific focus on LiDAR-based systems. We comprehensively explore the threat landscape, encompassing cyber-attacks on sensors and adversarial perturbations. Additionally, we investigate defensive strategies employed in countering these threats. This paper endeavors to present a concise overview of the challenges and advances in securing autonomous driving systems against adversarial threats, emphasizing the need for robust defenses to ensure safety and security.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to explore and summarize the adversarial robustness issues faced by LiDAR (Light Detection and Ranging) - based machine - learning perception systems in autonomous vehicles. Specifically, the paper focuses on the following aspects: 1. **The threat of adversarial attacks**: - Sensors (such as LiDAR, cameras, etc.) and machine - learning models in autonomous driving systems are vulnerable to adversarial attacks. These attacks include but are not limited to traditional Adversarial Machine Learning (AML) attacks and cyber - attacks against sensors. - Adversarial attacks cause misjudgments in machine - learning models by introducing carefully designed perturbations or making slight modifications to input data, thus affecting the safety and reliability of autonomous driving systems. 2. **Research on defense strategies**: - The paper investigates existing defense strategies to deal with the above - mentioned adversarial attacks. These defense strategies aim to improve the robustness of autonomous driving systems and ensure their normal operation in the face of malicious attacks. - Analyze the limitations of existing defense strategies and point out the deficiencies in current research, providing directions for future research. 3. **Technical review and challenges**: - The paper details the working principle of LiDAR sensors and their applications in autonomous driving, including point - cloud data processing, 3D object detection, semantic segmentation and other tasks. - Explore the characteristics of LiDAR point - cloud data (such as sparsity, noise, incompleteness, etc.), and analyze the impact of these characteristics on machine - learning models. - Emphasize the necessity of ensuring the robustness of machine - learning perception modules in autonomous driving systems, especially in the face of adversarial attacks. 4. **Contributions and prospects**: - Provide a comprehensive review of machine - learning models based on 3D LiDAR, especially those widely recognized and highly cited models. - Conduct in - depth analysis of adversarial attacks and defense strategies for 3D autonomous driving systems, revealing the gaps and emerging trends in current research. - Emphasize the importance of identifying and analyzing the limitations of existing AML defense strategies, providing a reference for the design of future defense mechanisms. In general, this paper is committed to filling the gaps in the existing literature, providing a comprehensive overview of the robustness research of LiDAR - based autonomous driving systems under adversarial attacks, and emphasizing the urgency of ensuring the safety and reliability of autonomous driving systems.

A Survey on Adversarial Robustness of LiDAR-based Machine Learning Perception in Autonomous Vehicles

Adversarial Robustness Analysis of LiDAR-included Models in Autonomous Driving

Navigating Threats: A Survey of Physical Adversarial Attacks on LiDAR Perception Systems in Autonomous Vehicles

Adversarial Sensor Attack on LiDAR-based Perception in Autonomous Driving

Robust Roadside Physical Adversarial Attack Against Deep Learning in Lidar Perception Modules

Towards Robust LiDAR-based Perception in Autonomous Driving: General Black-box Adversarial Sensor Attack and Countermeasures

Adversarial Computer Vision Via Acoustic Manipulation of Camera Sensors

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

PLA-LiDAR: Physical Laser Attacks Against LiDAR-based 3D Object Detection in Autonomous Vehicle.

Adversarial Objects Against LiDAR-Based Autonomous Driving Systems

Exploring Adversarial Robustness of LiDAR-Camera Fusion Model in Autonomous Driving

Exploring Adversarial Robustness of LiDAR Semantic Segmentation in Autonomous Driving

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

Deep Learning-Based Autonomous Driving Systems: A Survey of Attacks and Defenses

A Survey and Evaluation of Adversarial Attacks for Object Detection

Attacking vision-based perception in end-to-end autonomous driving models

A Survey of Robustness and Safety of 2D and 3D Deep Learning Models Against Adversarial Attacks

Adversary ML Resilience in Autonomous Driving Through Human Centered Perception Mechanisms

A Comprehensive Study of the Robustness for LiDAR-Based 3D Object Detectors Against Adversarial Attacks

Emerging Threats in Deep Learning-Based Autonomous Driving: A Comprehensive Survey

Evaluating Adversarial Attacks on Driving Safety in Vision-Based Autonomous Vehicles