Machine Learning Algorithm of Detection of DOS Attacks on an Automotive Telematic Unit

Today vehicles are connected to private networks which are owned by the car manufacturer. But in coming years, vehicles become more and more connected to the public Internet for infotainment applications but also to safety applications. Like any Internet terminal, some hackers can attack the wireless connectivity unit of the vehicle with Distribution Denial of Services (DDOS) attacks, so that the wireless connectivity unit of the vehicle is not available and the service is lost. Therefore, it is critical to developing a mechanism to detect such an attack and eliminate it, to maintain the availability of the wireless connectivity unit. This paper proposes an algorithm which proceeds in 2 steps: it uses an unsupervised machine learning algorithm to detect DDOS attacks in the incoming Internet data. When it detects an attack, it uses the results of the machine learning algorithm to split the legitimate flow and the rogue flows. The rogue flow is filtered so that the availability of the wireless connectivity unit of the vehicle is restored. This proposed algorithm needs very few CPU computing power and is compatible with low-cost CPUs which are used in an automotive wireless connectivity unit.