A Multi-Layer Fault Triggering Framework based on Evolutionary Strategy Guided Symbolic Execution for Automated Test Case Generation

Zhiyu Duan, Yujia Li, Pubo Ma, Xiaodong Gou, Shunkun Yang

Abstract: The powerful technique, symbolic execution, has become a promising approach for analyzing deep complex software failure modes recently. However, as the software scale grows rapidly in intelligent automatic control system, these methods unavoidably suffer the curse of path explosion and low global coverage. To solve the problem, an evolutionary strategy guided symbolic execution framework is proposed for triggering hard-to-excite input-relevant faults. A novel alternate asynchronous search strategy is adopted to enhance the breadth-search capability of symbol execution. Furthermore, by combining ANGR, a popular symbolic execution engine, and genetic algorithm, this method synchronously triggers the potentially hidden hybrid fault modes at different levels in the software architecture. Case studies on the SIR test suite demonstrate that the GA-enhanced symbolic execution greatly improves coverage and accelerates test convergence. Among them, the coverage rate has increased by up to 23.7%. With a baseline of 95% line coverage, the proposed method can reduce the number of iterations by at least 43.3%.