Satisfiability modulo fuzzing: a synergistic combination of SMT solving and fuzzing

Programming languages and software engineering tools routinely encounter components that are difficult to reason on via formal techniques or whose semantics not even available—third-party libraries, inline assembly code, SIMD instructions, system calls, calls machine learning models, etc. However, often access these is available as input-output oracles—interfaces query certain inputs receive the respective outputs. We refer such functions closed-box . Regular SMT solvers unable handle functions. propose Sādhak, a solver for theories modulo Our core idea use synergistic combination of fuzzer an engine solve constraints pertaining theories. The fuzz engines attempt converge model by exchanging rich set interface relevant interpretable them. implementation, demonstrates significant advantage over only other capable handling constraints: Sādhak solves 36.45% more benchmarks than best-performing mode this state-of-the-art has 5.72x better PAR-2 score; solved both tools, (on average) 14.62x faster.