Elimination by Linear Association: An Effective and Efficient Static Mirror Adaptive Random Testing

Adaptive random testing (ART) is a software testing method which combines randomness with even distribution of test cases within the input domain of a program with the aim of improving the effectiveness of random testing (RT). It was established right from the onset that, ART is considerably less efficient compared to RT due to the overhead cost involved in filtering randomly generated test cases in order to achieve the even spread objective. Again, it has been observed that over-concentration on achieving better effectiveness at the expense of efficiency will make ART advantage over RT a superficial one. Besides, the ART is close to its theoretical bound in terms of effectiveness. Various algorithms have therefore emerged that seeks to minimize the efficiency deficit incurred by the ART. One of such strategies is mirror adaptive random testing (MART). Unfortunately, the MART's performance is generally unstable due to the lack of diversity in mirror generated test cases. The culprit has been identified as the mirroring functions used in place of complex ART computations. In this paper, we present elimination (E) by linear association (E-MART) as a solution to the problem of the MART that guarantees diversity in all dimension(s) of mirror test cases. By partitioning the source domain into multiple subdomains, we systematically isolate mirror partitions which are linearly associated with the source domains. The source domain is then iteratively partitioned whiles forgetting strategy is applied to select test cases. The simulations and experimental studies conducted indicate that the EMART has a more stable performance compared to the MART and compares favorably in terms of efficiency by reducing the quadratic time of the MART to linear.