CAN-D: A Modular Four-Step Pipeline for Comprehensively Decoding Controller Area Network Data

Controller area networks (CANs) are a broadcast protocol for real-time communication of critical vehicle subsystems. Original equipment manufacturers passenger vehicles hold secret their mappings CAN data to signals, and these definitions vary according make, model, year. Without mappings, the wealth information hidden in packets is uninterpretable, severely impeding vehicle-related research, including cybersecurity privacy studies, aftermarket tuning, efficiency performance monitoring, fault diagnosis name few. Guided by four-part signal definition, we present CAN-D (CAN-Decoder), modular, four-step pipeline identifying each signal's boundaries (start bit length), endianness (byte ordering), signedness (bit-to-integer encoding), leveraging diagnostic standards, augmenting subset extracted signals with meaningful, physical interpretation. En route CAN-D, provide comprehensive review reverse engineering research. All previous methods ignore signedness, rendering them incapable decoding many standard definitions. Incorporating grows search space from 128 4.72E21 tokenizations introduces web changing dependencies. In response, formulate, formally analyze, an efficient solution optimization problem, allowing identification optimal set byte orderings. addition, two novel, state-of-the-art boundary classifiers—both which superior approaches precision recall three different test scenarios—and first classification algorithm, exhibits $>$ 97% F-score. Overall, only potential extract any that also state art. evaluation on 10 makes, CAN-D's average notation="LaTeX">$\ell ^1$ error five times better (81% less) than all lower error, even when considering meet prior methods’ assumptions. Finally, implemented lightweight hardware, on-board (OBD-II) plugin in-vehicle decoding.