Application Of Machine Learning Models In Vulnerability Mining Of Network Protocols

With the development of society,mankind is becoming more and more dependent on the Internet.The eternal blue virus,which is based on the vulnerability of Windows network sharing protocol in 2017,is wreaking havoc around the world.In a network or software system,if there are vulnerabilities that can be exploited,the risks to social security and personal security are enormous.So vulnerability mining technology has become an important research problem in information security field.The traditional method of exploiting vulnerability can not cope with the complex network security situation.In recent years,with the rapid devel-opment of artificial intelligence technology and machine learning tech-nology,it has solved many important problems in science,education,medical and other fields.In this paper,the fuzzy test(fuzzing)method is used to exploit the network protocol.Fuzzy testing techniques are used to exploit vulnerabil-ities by sending large amounts of abnormal data to test targets and moni-toring whether the software system is working properly.In this paper,we first need to analyze the protocol format and model,and use the fuzzy probe value to change the different parts of the protocol to generate a large number of test cases.These test cases are then sent to the test targets,and the network state and process state of the test target is monitored in real time.Finally,if the test case triggers the vulnerability,the system au-tomatically records the test case information and the vulnerability infor-mation.The Fuzzing algorithm can generate a large number of test cases,but the effectiveness of test cases is usually poor.Send a large number of test cases to the test target and monitor whether it has triggered the bug.The above process usually takes up a lot of time.This paper further improved the fuzzing algorithm and applied the machine learning model to fuzzing algorithm.The machine learning model can evaluate the large number of test cases generated by the fuzzing algorithm,and evaluate the results as test cases and invalid test cases that are easy to trigger vulnerabilities.The test case of machine learning evaluation as an easy to trigger vulnerability is sent to the test target,which can greatly reduce the time of exploit and improve the efficiency of exploit.At the same time,the system can study the structure and weight of the machine learning model,and make a key variation of the fields that are prone to the vulnerability.In this paper,we have exploited the vulnerability of IMAP and HTTP.We discovered the flaw of the mail server.At the same time,the test cases of fuzzing are evaluated with support vector machine algorithm and neural network algorithm,and the accuracy of evaluation is obtained.