4.5.2. Discussion

In this study, we propose an innovative bit-level reverse framework for automotive CAN messages. This framework builds a multiple linear regression model between CAN traces and sensor data, uses decision coefficients to filter candidate messages, and uses model parameters to determine how data fields represent vehicle behavior and maximally recover DBC files. In the test vehicle, this framework has high accuracy in both message screening and bit-inversion. However, the limitation of the test environment results in the unavailability of the extreme vehicle behavior data, leading to less than perfect results in bit-reversion. In addition, the framework reverses the candidate messages correctly in a short time, which improves the reversal efficiency. Our study proposes the only CAN message translator that can achieve bit-level reversal and has significant advantages over other existing methods for boundary delineation and message verification. Finally, the framework can be applied to any standard-compliant commercially available vehicle.
