Countering Adversarial Attacks on Autonomous Vehicles Using Denoising Techniques: A Review

Abstract

The evolution of automotive technology will eventually permit the automated driving system on the vehicle to handle all circumstances. Human occupants will be just passengers. This poses security issues that need to be addressed. This paper has two aims. The first one investigates strategies for robustifying scene analysis of adversarial road scenes. A taxonomy of the defense mechanisms for countering adversarial perturbations is initially presented, classifying those mechanisms in three major categories: those that modify the data, those that propose adding extra models, and those that focus on modifying the models deployed for scene analysis. Motivated by the limited number of surveys in the first category, we further analyze the approaches that utilize input transformation operations as countermeasures, further classifying them in supervised and unsupervised methods and highlighting both their strengths and weaknesses. The second aim of this paper is to publish CarlaScenes dataset produced using the CARLA simulator. An extensive evaluation study, on CarlaScenes, is performed testing the supervised deep learning approaches that have been either proposed for image restoration or adversarial noise removal. The study presents insights on the robustness of the aforementioned approaches in mitigating adversarial attacks in scene analysis operations.INDEX TERMS Countering adversarial images, robust road scene analysis, deep learning.

Securing the Future: A Comprehensive Review of Security Challenges and Solutions in Advanced Driver Assistance Systems

“…While adversarial instances for 2D pictures and CNNs have received a great deal of study, 3D data like point clouds have received less attention [142]. Investigating the influence of adversarial point clouds on established deep 3D models assumes paramount importance in numerous safety-critical 3D applications, notably in the domain of autonomous driving [143]. A key component of the attack method against 3D models is the creation of 3D adversarial point clouds.…”

Toward Autonomous Physical Security Defenses Using Machine Learning

“…Adversarial machine learning refers to a set of attacks techniques against intelligent systems, wherein the adversary aims to perturb, poison, or seal the underlying machine learning models [28,29]. These types of attacks are of particular concern in safety-critical applications [30]. This section explores the use of this mechanism to undermine the operations of the proposed tamper-proof detection mechanism and develops a countermeasure to enhance the security of our solution.…”

Robustifying cooperative awareness in autonomous vehicles through local information diffusion

“…To do so, scene analysis and understanding, i.e., visually detect nearby objects (e.g., vehicles, cyclists, pedestrians) by drawing bounding boxes, are necessary pre-processing steps. Deep learning can be considered paramount for this challenging perception task [13]. Localization is a step beyond perception, since it is required for optimizing the driving performance of vehicle through advanced path planning and control.…”