Research On Signal Extraction And Analysis Of In-Vehicle CAN Data Frame

With the development of automotive electronics and intelligence,the communication between electronic components in the vehicle is becoming more and more complex.The CAN bus is an important communication method between various electronic components in the vehicle.Vehicle manufacturers strictly maintain secrecy of the protocol format of signals in the CAN data frame.The protocol format of signals in the CAN data frame is widely used in the vehicle aftermarket,penetration testing,intrusion detection,and reverse engineering.Existing research works designed signal extraction methods based on CAN data frame messages,but did not consider the byte encoding format of CAN data frames.The multi-byte signal of the little byte endian encoding format was incorrectly divided into multiple signals,and the extracted signal lacks further analysis,the physical meaning of the signal cannot be marked.Therefore,this article designs a set of unsupervised in-vehicle CAN network signal extraction and analysis system,which automatically completes the estimation of the byte encoding format of the CAN data frame,the bit segmentation of the signal,the classification of the signal and the clustering of similar signals.The main work of this paper is as follows:1.The design and development of the data acquisition terminal.We clarify the data acquisition requirements in the vehicle,design and develop a data acquisition terminal which is based on wireless transmission and supports multiple acquisition of CAN,analog signals and level signals.Then we use the data acquisition terminal to collect the actual vehicle CAN data,and complete the construction of the data set combined with open source data.2.The design and implement of unsupervised signal extraction algorithm for CAN data frame.According to CAN data frame signal coding specifications and coding characteristics,combined with greedy search,an unsupervised CAN data frame signal extraction algorithm is designed.Experimental results show that the algorithm can identify the encoding format of CAN data frames and effectively improve the accuracy of signal bit segmentation.3.The design and implement of signal classification and clustering algorithm for CAN data frame.After extracting the signal,the signal classification algorithm is designed according to the coding characteristics of each type of signals.And then,the continuous physical signal is hierarchically clustered based on the cosine similarity.The correlation between signals is visually displayed,which is helpful for marking the physical meaning of signals.4.Integrated platform development and testing.We design and develop an integrated software platform for CAN data frame acquisition and signal extraction.The platform has the ability of offline signal extraction,online data acquisition,signal analysis and waveform drawing.The data acquisition terminal and the integrated software platform are tested as a whole.The test results verify that the functions and performance of the data acquisition terminal and the integrated software platform meet the design requirements.