Methods for improving Fuzzing-Testing Using Machine Learning and visualisation of results

The article is devoted to the analysis of fuzzing testing - a method of dynamic testing of a program's binary code. The analysis of the literature allows us to assert that, today, automated fuzzing testing is quite complex in terms of building algorithms, as well as an extremely demanded process from the point of view of information security. The implementation of this method for analyzing the dynamic binary code of a program using machine learning is the most preferable, since it implies a fairly thorough work with data, namely: analysis of the input data of the program, mutation (modification) of data, analysis of reports on the program's abnormal termination. The conducted research allows us to conclude that robustness is a necessary property of modern software. The classification of the main types of binary vulnerabilities has been carried out. © 2021 IEEE