Automated Vehicle Electronic Control Unit (ECU) Sensor Location Using Feature-Vector Based Comparisons
Gregory S. Buthker · OhioLink ETD Center (Ohio Library and Information Network) · 2019
In the growing world of cybersecurity, being able to map and analyze how software and hardware interact is key to understanding and protecting critical embedded systems like the Engine Control Unit (ECU). The aim of our research is to use our understanding of the ECU's control flow attained through manual analysis to automatically map and identify sensor functions found within the ECU. We seek to do this by generating unique sets of feature vectors for every function within the binary file of a car ECU, and then using those feature sets to locate functions within each binary similar to their corresponding control function. This feature algorithm is used to locate candidate functions that utilize a sensor, and then examine the structure of each of these candidate functions to approximate the memory-mapped IO address of each sensor. This method was able to successfully locate 95\% of all candidate functions and was able to successfully recover 100\% of likely sensor addresses within each of those functions.