Constructing subtle higher order mutants for Java and AspectJ programs

One goal of higher order mutation testing is to produce higher order mutants (HOMs) that represent subtle faults. We define subtle HOMs as those that are not killed by an existing test set that kills all the first order mutants of a given program. The fault detection effectiveness of the test set can be improved by adding test cases that kill subtle HOMs. However, finding subtle HOMs can be costly even for small programs because of the large space of candidate HOMs. Moreover, a large majority of HOMs are killed by test sets that kill all first order mutants, making the subtle ones relatively rare. We introduce three search-based algorithms (Genetic Algo-rithm, Local Search, and Random Search) for finding subtle HOMs in Java and AspectJ programs. All three algorithms found subtle HOMs for all studied programs but Local Search was more successful in finding subtle HOMs than Genetic Algorithm and Random Search.