Intelligent hyper heurisitic algorithms to automate test data generation for critical software applications testing

Abstract

newline This work aims to develop a generalized, extensible framework for automated software application test case generation. The algorithms of this research work shall generate test cases satisfying multiple criteria in lesser iterations and be effective in fault detection. Two kinds of problems are identified: one based on gray box and another based on black box testing. For the first problem, the three subject programs were taken from the safety-critical domain. The coverage criteria considered are Data Flow Coverage or Def-Use paths Coverage and Software Safety Requirements as the primary criteria, and Combinatorial Interaction Testing coverage on Equivalence Partitioning as sub-criteria. Experiments were conducted with existing algorithms in the literature, considering one criterion at a time. Genetic Algorithm (GA) for data flow coverage and Random testing for Combinatorial Interaction testing were implemented, and the coverage criteria were studied. The uncovered Def-Use paths by GA were analyzed. The following limitations were seen: variables in the def and use path were in True and False paths of the same conditional check, which can be exercised by a test sequence in different iterations of the program but not in a single iteration; GA was oscillating for floating point equality/inequality check.