Discerning the operational state of a vehicle’s distributed electronic systems from vehicle network traffic for use as a fault detection and diagnosis tool

This paper suggests a novel approach to finding faults in a vehicle’s electronic systems by monitoring the network traffic directly and generating statistical traits. The nature of the data in a CAN network is considered, and a case for the use of statistical analysis presented. Statistical traits extracted from the temporal behavior of network messages are investigated as a metric for fault detection. It is shown how this trait information can be extracted from network data, and how this information could be used for fault detection of an unknown fault on a CAN network. It is then demonstrated that combining multiple types of trait data can be used to correctly identify a fault once detected