Effects of the Sampling Technique on Sender Identification Systems for the Controller Area Network

Abstract

As a result of the ongoing development of vehicle electronics and additional wireless communication interfaces, the possibilities for attacks and their negative consequences are increasing. Once an attacker has obtained access to the internal vehicle communication, in the case of the Controller Area Network (CAN) the attacker is able to forge all messages of the connected Electronic Control Units (ECUs) without a receiving ECU being able to recognize any suspicious behavior. The use of cryptographic methods is only possible to a limited extent due to restricted resources of the ECUs, which is why sender identification systems have been presented which are able to detect these kind of attacks. Presented approaches use different procedures to capture the analog signals on which the detection of attacks respectively the identification of the sender is based. This work shows that the impact on the performance of the sender identification system by the different sampling methods is minimal and therefore the selection of the appropriate technique can be mainly based on the available resources and the communication structure of the corresponding vehicle platform. This is shown on the one hand by the direct analysis of the analog signals captured from a real vehicle as well as by an evaluation of the previously introduced sampling methods using a recently published sender identification system. In addition, an assessment of the procedures based on different parameters shows which method is to be preferred for which application.