Intelligent Fuzzy Attack Vector Generation Framework Based on Deep Adversarial Learning

With the advancement and proliferation of Internet of Things (IoT) technology, IoT has become intimately linked with various aspects of human life and production, especially in the realm of industrial control. And the security of industrial control protocols is imperative to ensure the security of industrial control systems. Fuzz testing has been applied to protocol testing, but traditional fuzzing methods require researchers to have a clear understanding of the specific protocol specifications and details of their implementations in order to conduct effective testing. This approach demands a considerable investment of time, resources, and effort. Therefore, we propose an industrial control protocol fuzz testing framework based on adversarial generative networks. The core module of this framework, the adversarial generation module, mainly employs an improved GANs network to train the model to learn the syntax of the protocol. The trained network generates fake but credible test protocol messages. This framework addresses the issue of high costs associated with traditional fuzz testing methods. Other modules will be introduced in Section 3. This paper tests a common protocol, Modbus. The experiments demonstrate that our framework can rapidly generate a large number of fake but credible test protocol messages, enabling efficient fuzz testing of the protocols.