Research on Test Cases Screening Method Based on Generative Adversarial Networks
Zhongwei Li, Ji Sun, Xianji Jin, Zihan Ma · 2024
In order to overcome the limitations of traditional fuzzy testing methods in generating diversified and high-coverage test cases, this paper adopts the technique of generative adversarial networks (GANs) and proposes a test case screening method based on an improved Information Maximizing Generative Adversarial Network (infoGAN). To improve the efficiency and effectiveness of fuzzy testing of industrial robotic systems, the adversarial training of generators and discriminators is optimized, enabling the improved infoGAN model to generate high-quality and diverse test cases. The test results indicate that the test cases screened by this method significantly enhance the coverage and effectiveness of vulnerability detection in industrial robots and improve the overall impact of fuzzy testing.