Phase One: Learning and Imitation

By collecting massive amounts of normal CAN communication data, the AI model learns the statistical distribution and structural features of the protocol, becoming a "protocol expert." This enables it to generate test cases with a format compliance rate exceeding 87%, successfully penetrating the first layer of format validation.

Phase 2: Injection and Mutation

Building upon the ability to generate "normal data," perform targeted perturbations within the latent space to create messages that "appear normal but hide secrets," such as tampering with critical data segments to boundary values, thereby crafting "wolves in sheep's clothing"—high-value test cases with correct formatting.

Phase 3: Closed-Loop Testing

Continuously inject intelligent test cases into the system, monitor bus load, ECU status, and vehicle behavior, forming a closed loop of "attack-feedback-optimization."

Protocol-Aware Intelligent Mutation

The core of the ASDL CAN Fuzzer is the "Protocol-Aware" approach. It does not blindly perform random mutations but designs tests based on a deep understanding of protocol specifications such as CAN, CAN FD, UDS, and ISO-TP. This shares the same underlying principle as DCGAN's "learning protocol patterns," both significantly reducing meaningless data and focusing resources on effective attacks.

Comprehensive and efficient test coverage

This tool systematically traverses normal and abnormal values of protocol fields (such as ID, DLC, RTR, Payload) using "fingerprint recognition" and a combination model. With a vast number of targeted test cases (in the hundreds of thousands), it outperforms industry-renowned tools like Defensics in deep protocol coverage, including CAN 29-bit ID, RTR bit, and ISO-TP large frame transmission, ensuring both the breadth and depth of vulnerability discovery.

Engineering-friendly automation process

The ASDL CAN Fuzzer establishes a complete automated workflow from Test Plan → Interoperation → Test Case → Execute → Result. Users can quickly set up a test environment by importing DBC/ARXML files and configure complex test sequences using intuitive XML templates, significantly lowering the high technical barrier of traditional Fuzz testing.

Collaborative Design for Teams

The tool offers an enterprise-level license, supporting up to 5 users simultaneously, which addresses the bottleneck of a single license. It is highly suitable for R&D team collaboration, enhancing overall testing efficiency.