4.4.3. Execution Complexity

We compare the algorithms in this section concerning the devices needed for their execution, the data requirement, the algorithm execution time, and the reverse results. As shown in Table 12, each algorithm relies on OBD-II data acquisition devices. Only the framework and LibreCAN require additional sensor devices and smartphones, respectively. In terms of data requirements, the READ and ReCAN require only CAN traffic, the linear regression method and LibreCAN require data from additional devices. However, the correlation coefficient method requires UDS data through interaction with the vehicle [47]. LibreCAN is the algorithm that takes the longest time to execute since some manual work is also required, and the fastest execution is the correlation coefficient method. The framework in this paper is close to the average time of the READ algorithm. However, in terms of reverse results, our scheme is the only one that can achieve bit-level reverse, outperforms the other algorithms in boundary delineation and message filtering, and does not require interaction with the vehicle. Although additional sensor devices are required, such sensors can be purchased very cheaply and used very simply in the market.

**Table 12.** Execution complexity comparison of different algorithms.


*4.5. Application and Discussion*

4.5.1. Application

The bit-level automotive CAN bus reverse framework proposed in this study can be used in almost all commercially available vehicles, independent of vehicle make and model. According to Table 12, the implementation of the framework requires OBD-II [48,49] data collection devices, sensors, and CAN traffic. In-vehicle CAN network traffic is typically collected using the OBD-II interface, a globally accepted automotive standard. It is required for almost all commercially available vehicles to be equipped with an OBD-II interface before they can be marketed [50–54]. Therefore, regardless of vehicle models on the market, the vehicle CAN traffic can be obtained after connecting OBD-II data collection devices. Therefore, regardless of vehicle models on the market, the vehicle CAN traffic can be obtained after connecting OBD-II data collection devices. For OBD-II data acquisition devices, such devices are readily available on the market today, with prices ranging from a few tens to a few hundred dollars. The sensor devices used in this framework are off-the-shelf motion sensors, which are inexpensive and easily placed in various vehicle parts to collect relevant data. Using CAN traffic and sensor data as input to our proposed

framework, the algorithm proposed in this paper can obtain how CAN messages in any vehicle describe the vehicle state.

To verify the applicability of the framework, an electric car with completely a different power and brand was chosen to apply the framework. The reverse results are shown in Table A1 in Appendix A. In the absence of relevant DBC files, a script is provided in the appendix that can display CAN data changes in real-time to confirm the accuracy of each result. All filtered messages are consistent with the actual results in the actual results, and the reverse results of the bits remain consistent with the data bit changes. Overall, the method proposed in this study can be applied to most vehicle CAN message inversions and is not affected by vehicle changes.
