A Fuzzing Data Generation Method Based on Conditional Generative Adversarial Networks

Testing network protocols is a challenging task, and fuzzing, as one of the mainstream methods in this field, has demonstrated its practical value and validity. However, the complex syntax and state transition of network protocols impose limitations on existing approaches when testing stateful and proprietary protocols, significantly impacting their applicability and efficiency. In this paper, we propose a testing case generation method guided by real-time message traffic, which can not only learn the format of protocol messages and reconstruct semi-valid test cases that conform to the syntax but also respond correctly based on the current state of the protocol. Experimental results demonstrate that our approach provides more effective support for industrial network protocol fuzz testing, compared with several existing methods.