Research And Implementation Of Intelligent Protocol Fuzzy Testing Technology Based On Deep Learning

With the development of the Internet,the network has become an indispensable part of people’s daily life and work,but it also brings many network security problems.The complexity and vulnerability of network protocols are one of the main causes of network security problems.Among the vulnerability detection methods for network protocols,fuzzy testing can effectively identify security vulnerabilities by providing a large number of unexpected inputs to the target program.An important part of fuzzy testing is the generation of fuzzy test samples.Traditional fuzzy testing tools generate low diversity of test samples and low acceptance rate of test inputs,resulting in poor vulnerability detection and low generality.Although existing fuzzy testing methods based on deep learning can effectively learn protocol features,they often suffer from unstable model training,disappearing gradients,and difficulty in generating discrete protocol test samples.To address the above problems,this paper designs a deep learning-based protocol fuzzy test sample generation model to generate fuzzy test samples by automatically learning the format features of the original protocol through the model,and designs a corresponding intelligent protocol fuzzy test framework.The main research contents of this paper are as follows:1.A protocol fuzzy test sample generation model based on an improved sequence generation adversarial network is proposed.Aiming at the problems of unstable model training and gradient disappearance of existing sample generation methods,this model can automatically learn the spatio-temporal characteristics of protocol data.The model designed in this paper includes two parts:the generative model and the discriminative model.Among them,the core of the generative model is a convolutional long and short term memory network,and the discriminative model is a convolutional neural network.The generative model and the discriminative model are trained to reach the equilibrium state by cyclic game training.The trained model is used to generate protocol fuzzy test samples so as to perform fuzzy testing.The results show that the proposed method achieves a higher acceptance rate of test inputs and is more efficient compared with existing methods.2.To address the problems of low vulnerability detection capability and poor generality of existing fuzzy testing tools,this paper proposes an intelligent protocol fuzzy testing framework based on deep learning,which is divided into protocol data collection module,protocol data preprocessing module,protocol fuzzy test sample generation module,protocol fuzzy test module and monitoring and anomaly analysis module.Experiments show that the framework generates an average of 5 vulnerabilities per 10,000 samples,and the test sample pass rate reaches 78.1%,which is a significant improvement compared with existing methods.3.For Modbus-TCP protocol,the protocol fuzzy test sample generation model proposed in this paper and the designed protocol fuzzy test framework are verified functionally and analyzed performance.The test results show that the model can automatically learn the characteristics of protocol data and generate effective protocol fuzzy test samples,and the protocol fuzzy test framework designed in this paper can perform intelligent vulnerability detection for the target system.