Security Analysis of Vehicle Instrument Clusters by Automatic Fuzzing and Image Acquisition

Automatic testing and reverse engineering of in-vehicle components have become topics of significant interest as they allow industry professionals to get a better understanding of component behaviour and attack surfaces. In this work we pursue the automatic testing of in-vehicle instrument clusters. These devices may be subject to attacks that may mislead the driver regarding the current state of the car. Understanding the feasibility of such attacks requires the ability to determine the behaviour of the cluster in response to specific instructions received from the CAN bus. For this purpose we use fuzz testing and a camera in order to remove manual intervention and set room for automatic learning of the commands that can be sent to the cluster. This enables the automatic detection of the commands with little or no human intervention. We test our framework on three clusters from real world vehicles and determine in an automatic manner a sufficiently large number of commands and responses that they trigger.