Attacks on Self-Driving Cars and Their Countermeasures: A Survey

Abstract

Intelligent Traffic Systems (ITS) are currently evolving in the form of a cooperative ITS or connected vehicles. Both forms use the data communications between Vehicle-to-Vehicle (V2V), Vehicleto-Infrastructure (V2I/I2V) and other on-road entities, and are accelerating the adoption of self-driving cars. The development of cyber-physical systems containing advanced sensors, sub-systems, and smart driving assistance applications over the past decade is equipping unmanned aerial and road vehicles with autonomous decision-making capabilities. The level of autonomy depends upon the make-up and degree of sensor sophistication and the vehicle's operational applications. As a result, self-driving cars are being compromised perceived as a serious threat. Therefore, analyzing the threats and attacks on self-driving cars and ITSs, and their corresponding countermeasures to reduce those threats and attacks are needed. For this reason, some survey papers compiling potential attacks on VANETs, ITSs and self-driving cars, and their detection mechanisms are available in the current literature. However, up to our knowledge, they have not covered the real attacks already happened in self-driving cars. To bridge this research gap, in this paper, we analyze the attacks that already targeted self-driving cars and extensively present potential cyber-attacks and their impacts on those cars along with their vulnerabilities. For recently reported attacks, we describe the possible mitigation strategies taken by the manufacturers and governments. This survey includes recent works on how a self-driving car can ensure resilient operation even under ongoing cyber-attack. We also provide further research directions to improve the security issues associated with self-driving cars.

A Probabilistic Approach to Estimating Allowed SNR Values for Automotive LiDARs in “Smart Cities” under Various External Influences

“…The essence of the attack lies in the concealed perturbations in the points cloud (the input 3D data of the object) carried out by the attacker, which lead to incorrect detection and/or recognition of the objects, or even the inability to identify and/or classify them. This type of attack is also called adversarial machine learning, as it is aimed precisely at deceiving the algorithms and methods of machine learning used that are sensitive to distortions or artificial perturbations [22].…”

“…Lidar's ECU malware attack. Infection of the ECU via the malware can cause both DoS of the LiDAR (or the whole AV at once) and malicious disturbances in its functioning [22]. Such an attack can be carried out if the intruder has physical access to the AV (e.g., Evil Mechanic attack [11]) or through the adversary impact on the ECU (for example, when upgrading firmware; over-the-air attack) [38].…”

Design, Analysis and Evaluation of a Lightweight Symmetric Key Stream Generator for Smart Meter and Autonomous Vehicle Security

“…i.e., Period= (2 29 -1) x (2 23 -1) x (2 19 -1) 2 71 To increase the level of complexity for attackers on the key stream generator, multiple feedback polynomials have been included in the design. Thus, the period increases by 2 12 and avoids the issue of being insecure. Therefore, the period of the proposed key stream is approximately = [(2 29 -1) x (2 23 -1) x…”

“…To protect the vehicular data, encryption is a must. Abdullahi Chowdhury et al, [12] have presented a detailed survey on the security issues related to connected vehicles and autonomous vehicle technologies. In [13] Sahand Murad et al, has dealt the security of the vehicular data shared over cloud by means of fragmentation and encrypting.…”

Threats and Attacks to Modern Vehicles

“…In [4] is being discussed different attacks on authenticity, integrity and confidentiality. In this article the focus is more Table 1.…”