Intrusion detection system for on-board automotive networks

Onboard automotive calculators or ECU (Electronic Control Unit) is responsible for a growing number of features in the vehicle. These calculators create a true on-board network by exchanging data via communication buses to coordinate their actions. While historically this network could be considered a closed system, security researchers have proved the possibility of an attack on an on-board network. With the appearance of many means of communication in automobiles, this network is now open to the outside and raises many safety problems in this area. In this thesis, we will first present the characteristics of embedded systems in the automobile. Then we will study the different attack scenarios that can target the on-board ECU's of modern automobiles. Next, we will present an intrusion detection system for on-board automotive networks based on the observed network traffic. Then we will present the various existing or planned computer security mechanisms to protect on-board automotive networks. Finally we will implement a learning model in order to learn attack messages injected on a CAN bus.