Which type of exclusion region is better for restricted random testing?: an empirical study

As an enhancement of Random Testing (RT), Adaptive Random Testing (ART) has been well studied, which aims at achieving well-performing distribution of random test cases. Previous studies have indicated that ART can effectively enhance the failure-detection ability of RT, and many implementations of ART have been proposed based on different notations. Among them, Restricted Random Testing (RRT) is a popular version of ART, which generates test cases outside the exclusion regions (i.e., the constructed regions that are located around the executed but no failure-causing inputs). As we know, there are many factors that may influence the performance of RRT such as the size and shape of exclusion region. In this paper, we conducted a series of simulations to investigate the impact that the type of exclusion region has on the testing effectiveness of RRT, i.e., Which type of exclusion region is better for RRT? The results show that when the failure-causing inputs cluster into a few failure regions or a large predominant failure region exists within the input domain, the uniform exclusion region is a good choice for RRT. However, when the failure region is less compact, it appears to be less suitable, because the irregular exclusion region has much better performances.