Empirical Evaluation of the Effectiveness and Reliability of Software Testing Adequacy Criteria and Reference Test Systems

This PhD Thesis reports the results of experiments conducted to investigate the effectiveness and reliability of ‘adequacy criteria’ - criteria used by testers to determine when to stop testing. The research reported here is concerned with the empirical determination of the effectiveness and reliability of both tests sets that satisfy major general structural code coverage criteria and test sets crafted by experts for testing specific applications. We use automated test data generation and subset extraction techniques to generate multiple tests sets satisfying widely used coverage criteria (statement, branch and MC/DC coverage). The results show that confidence in the reliability of such criteria is misplaced. We also consider the fault-finding capabilities of three test suites created by the international community to serve to assure implementations of the Data Encryption Standard (a block cipher). We do this by means of mutation analysis. The results show that not all sets are mutation adequate but the test suites are generally highly effective. The block cipher implementations are also seen to be highly ‘testable’ (i.e. they do not mask faults).