Research On Fuzzy Testing Methods For Industrial Control Protocol

With the continuous development of the Internet,people’s research on the security of industrial control systems is becoming more and more in-depth.For Fuzzing of network protocols,there are many problems with traditional Fuzzing technology.For example,the Fuzzing process cannot be effectively developed without mastering the protocol specifications.If random variation occurs,Fuzzing will be inefficient.The Generative adversarial network(GAN)plays a unique role in data generation,but there are many shortcomings in the application of protocol Fuzzing,such as the traditional GAN can not deal with discrete data,model convergence is difficult,and so on.In addition,when the Fuzzing case set is obtained,how to improve the probability of detecting vulnerabilities in a limited time is also one of the difficulties.In order to improve the efficiency of industrial control protocol Fuzzing,the following research is carried out:(1)Design Fuzzing method for industrial control protocol based on improved ranking Generative adversarial network(Rank GAN).Aiming at the problem that traditional GAN can not deal with discrete data and the model sometimes can not converge,this paper proposes a method of Fuzzing case generation for industrial control protocol based on improved Rank GAN based on the Rank GAN network model.This method performs batch standardization on the generator of the original Rank GAN network,fixing the mean and variance of the input data,effectively avoiding the occurrence of extreme situations in the data distribution;Adding a self attention mechanism to Ranker can further improve the quality of generated data.After experiments,it has been proven that the model can quickly converge,and the resulting generator can generate a set of test cases with high acceptance rate and high diversity.(2)Design a test case sorting method based on Simple Recurrent Units(SRUs)and attention mechanism.Aiming at the problem that the traditional protocol Fuzzing method has long invalid execution time and low efficiency,this paper proposes a case ranking method for industrial control protocol Fuzzing based on SRU network and attention mechanism.This method sorts the test case set according to the scores calculated by the model by learning the data format of effective test cases,and prioritizes the test cases that are most likely to detect vulnerabilities.When the time requirement is strict or the number of test cases sent is limited,sequential execution of the top test cases can improve the vulnerability detection rate and further improve the efficiency of Fuzzing.(3)Design and implement a Fuzzing prototype system for industrial control protocols.Aiming at the problems of complicated operation and process of traditional industrial control protocol Fuzzing,this paper designs and implements a Fuzzing prototype system for industrial control protocol.The system can realize the whole process of Fuzzing of industrial control protocols.First,the test case set with high acceptance rate and high diversity of the corresponding protocol is obtained through research point 1,and then the test case set obtained is sorted according to the scoring results through research point 2.The overall efficiency of Fuzzing of industrial control protocols can be improved through case generation and sorting in turn.Moreover,the system is simple to operate,which provides convenience for testers to analyze the test results.