Research On Vulnerability Detection Of Industrial Control Protocols Based On Fuzz Testing Techniques

Industrial control protocol is an important part of industrial software,as the main information transmission specification of industrial software However,the wide variety of industrial control protocols increases the risk of vulnerabilities.Therefore,in recent years,the security risks of industrial control protocols have been widely recognized as a hot topic in industrial security.Given the open and diverse nature of industrial control protocols,how to conduct vulnerability detection quickly and effectively has become a research hotspot in the field of industrial security.Fuzzy testing is a commonly used approach for detecting vulnerabilities in industrial control protocols.However,traditional fuzzy testing methods have drawbacks such as poor performance,the need to analyze the format information of specific industrial protocols,and lack of generality,which makes them inadequate for testing numerous industrial protocols.(1)In existing fuzzy testing methods for industrial control protocols,the efficiency of test case generation is low,the fuzzy testing results are poor,and the generated test cases cannot accurately represent protocol format information.To address this problem,this thesis proposes two fuzzy test case generation methods based on reinforcement learning.These methods are the fuzzy test case generation model based on the sequence generative adversarial network(Seq GAN)and the fuzzy test case generation model based on inverse reinforcement learning(IRL).The Seq GAN model uses LSTM as a generator to generate industrial control protocol data,and CNN as a discriminator to make the generated data more similar to real industrial control data.The IRL method also uses LSTM as a generator to generate industrial control protocol data,MLP as a reward approximator to calculate the gradient between generated data and real data,and to provide feedback to the generator model to generate data that is closer to real data.This thesis presents two models that increase the method of generating fuzzy test cases.In order to verify the effectiveness of the proposed method in this thesis,simulation fuzzy testing experiments were conducted on the commonly used industrial protocol Modbus TCP.Through the experiments,it was demonstrated that the two fuzzy testing case models proposed in this thesis have an 11%increase in vulnerability detection ability and a 17% increase in test case pass rate compared to the original GAN network model and the traditional fuzzy testing tool Peach.Additionally,the two models proposed in this thesis have higher diversity in fuzzy testing cases compared to GAN.The Seq GAN model generates test cases with a higher pass rate,while the IRL model can generate more diverse fuzzy test cases.(2)The current process of fuzz testing for industrial control protocols is relatively complicated and lacks an overall fuzz testing application system.This paper designs and implements a fuzz testing system,protocol data preprocessing,test case generation,fuzzy testing interaction,model testing tasks,vulnerability analysis,and more.This system achieves an integrated fuzzy testing process,reduces the threshold for fuzzy testing of industrial protocols,and serves as a reference for automating fuzzy testing of industrial protocols and software.This paper conducted specific simulation experiments and fuzzy testing on Modbus protocol,MQTT protocol,and MMS protocol,and successfully detected four security vulnerabilities.The results also demonstrate that the proposed fuzzy testing system can detect abnormal situations of vulnerabilities in industrial control protocols.The main research object of this thesis is industrial control protocols.Two industrial control protocol fuzzing test case generation methods were constructed to generate simulation data for industrial control protocols.Through experiments,the superiority of these two methods over traditional generation methods was demonstrated.Based on these two test case generation methods,an industrial control protocol fuzzing test system was constructed and its feasibility was demonstrated through simulation experiments of multiple protocols.