Classification of Attacks On Autonomous Cars
Automotive manufacturers have stated that completely autonomous driving, referred to by Tesla Motors as FSD, will become a part of the market in the coming years, and by 2030. There are a lot of advantages once FSD comes out of beta into general availability. Driver-assist features have been made available for a long time and saved countless hours in driving. Such features also help people like the elderly and specially challenged groups of people who are impaired from one of their limbs and are unable to drive cars. However, there is another aspect that should not be completely ignored.
Nowadays cars are fast progressing towards completely self-driving themselves on roads, however, we also must be increasingly careful as we add more and more features to our cars. The more connected the vehicle infrastructure and onboard electronics are, the easier it becomes for someone with mala fide intentions to break in and gain access to exploit it to do things according to their own will.
The increased connectivity combined with autonomous driving functions poses a considerable threat to the vast socio-economic benefits promised by AVs. However, there is not much historical data available on autonomous driving which means that traditional methods of risk assessment become ineffective. Thus, the authors are trying to explore the security aspect of connected cars and autonomous driving technology. Through this research, they want to provide cybersecurity professionals and automakers with the tools and knowledge to identify vulnerabilities, exploits and even give recommendations for mitigating any threats to the car and onboard computing infrastructure. Anyone who is working towards making cars safer like policymakers, professionals can find our project to be extremely helpful to them for their research. The analysis has been conducted by using a prototype based on Reinforcement Learning (RL), Proximal Policy Optimisation (PPO), and Sim2Real learning.