A Software Framework for Jetson Nano to Detect Anomalies in CAN Data

The current pace of development of cyber-physical systems requires the elaboration of fast methods for analyzing data circulating in them. Anomalies are patterns of data that do not conform to the concept of normal (expected) behavior. The data available on vehicular CAN bus reflects the current state of a vehicle, and it is the product of the engine control system and various sensors. In the present paper, we present software to process data from the CAN bus with the goal to detect anomalies in it. Often the data circulated in a vehicle is vendor-specific, in addition, we consider various methods for finding anomalies, therefore, it is advisable to design extensible software in the form of a software framework. The work is intended for the Jetson Nano platform, but can be run on another embedded Linux platform with restrictions on detection methods. We discuss hardware and software methods to obtain information on current state of the vehicle in real time, and then we briefly study how to implement anomaly analysis methods on the received data. Evaluation of detection methods is not included in the goals of the work; we mainly provide infrastructural methods for receiving data from the bus, decoding it and passing it to an anomaly predictor. Software was implemented in C++ with the ability to run Python code for the prediction, tests were carried out on a Mazda 6 first generation car and its ECU.