Optimizing the Parameters of an Evolutionary Algorithm for Fuzzing and Test Data Generation

Communication interfaces are particularly challenging to test using automatically generated test data. The test data sent through the interface must be ”valid enough” to overcome initial sanity checks of the interface and reach functions deep inside the integrated software. Machine-readable information about what data forms ”valid enough” messages is rarely available to test data generation tools. So instead, we evolve the messages with an evolutionary algorithm. This enables efficient fuzz testing for the communication interface between a satellite and its ground station. In this paper, using an algorithm implementation in our fuzzing tool DCRTT, we investigate the impact of algorithm parameter selection on the performance and the possibility of efficient general default parameter values. The preliminary results promise significant improvements to automated testing with respect to software security testing and quality assurance.