A Variational Autoencoding Method for Fuzzing Test Case Generation

Abstract

The invention discloses a variational self-encoding fuzzy test test case generation method, comprising the following steps: A. For different test objects, collect communication data conforming to the test object protocol format as training samples; B. Pass the VAE encoder , extract the feature space of the training data samples; C. Propose three generation strategies, and generate the corresponding feature space according to the three data generation strategies; D. According to the generated new feature space, through the VAE decoder, generate a new test case. The method can be applied to a variety of different industrial control network protocols, rather than for a certain protocol. Its biggest feature is that it learns the characteristics of sample data through machine learning, and generates new test cases based on the characteristics and VAE technology, which avoids the need for testers to be proficient in protocol-related knowledge when generating test cases from traditional fuzzing tests, which greatly reduces the number of tests. The burden on personnel increases the efficiency and scope of testing.

Description

technical field [0001] The invention relates to a method for generating a fuzzy test test case, in particular to a method for generating a fuzzy test test case by variational self-coding, and belongs to the technical field of test case generation methods. Background technique [0002] Fuzz testing generates a large amount of abnormal data and sends the data to the test object, expecting to find data that can cause the test object to work abnormally. Through detection technology, abnormal data records are analyzed to discover vulnerabilities hidden in industrial networks. According to the definition of fuzz testing, its specific steps can be divided into the following three steps, such as figure 1 shown. Fuzz testing is a gray-box test, which does not need to obtain all the source code of the test object in advance, but needs to generate abnormal data that can be accepted by the industrial control network device, because the industrial control network device will directly d...

AI Technical Summary

Problems solved by technology

The data generated based on the generated test case generation method can be directly received by the test object and has a higher acceptance rate, so it is a hot spot in most researches at present. However, this method requires researchers to be proficient in protocol-related knowledge, which greatly increases The test time and difficulty of fuzz testing are reduced, and for private protocols whose protocol formats are not disclosed, test cases cannot be generated

Images