Optimizing multi-input scheduling for fuzzing test using Bayesian algorithm weight optimization

As a major software testing method, fuzzing has the potential to enhance testing coverage and detect a great number of software vulnerabilities in complex code programs. The current fuzzy testing for multiple input parameters lacks an effective calling mechanism and treats all variant parameter equally, resulting in low efficiency. This paper presents a fuzzy testing method based on Bayesian algorithm for scheduling parameters. Firstly, it combines the characteristics of input parameters to match the appropriate variation strategy. Secondly, it utlizes the Bayesian algorithm to automatically generate weight values for the objective function index during sufficient testing. It evaluates test cases with multi-parameter inputs according to these weights and selects parameters with high utilization value for multiple tests. Finally, in order to evaluate the effectiveness of the method of this paper, an application programming interface of Amercian fuzzy lop (AFL-API) is developed based on the typical fuzzy testing tool AFL++, and compared with AFL++ as a fuzzy testing tool. The experimental results demonstate that when using the AFL-API for program testing, there was 47% increase in coverage rate, 52% increase in the number of discovered crashes, and more program vulnerabilities were identified during actual testing.