Input Invariants in Fuzz-testing

Fuzz-testing is a technique in which test inputs are generated programmatically to enhance software testing efficiency. This study investigates the suitability of ISLa, a declarative specification language to improve fuzz-testing. ISLa augments context-free grammars (CFG) with additional constraints to express context-sensitive input invariants. In the present study, a testing setup is developed, in which ISLa is used to specify invariants of valid test inputs, as well as test-case-specific preconditions. The expressiveness of ISLa as a specification language is evaluated, as well as the ISLa's effectiveness in generating test inputs which reveal implementation bugs. Ultimately, it is concluded that while ISLa helps tackle some challenges in generating test input, it has some fundamental and practical limitations that prevent it from being widely applicable as an input generation tool.