fuzzing

Fuzzing, or fuzz testing, is an automated software testing technique that feeds invalid, unexpected, or random inputs to a computer program in order to trigger crashes, assertion failures, or other abnormal behavior. The goal is to uncover vulnerabilities and robustness problems that arise from improper input handling, memory safety issues, or logic errors. Fuzzing is widely used in software security to identify bugs that could be exploited in real-world attacks, as well as to improve input validation and fault tolerance.

The technique originated in the late 1980s with experiments by Barton Miller at the University of California, Berkeley, who demonstrated that random inputs could cause surprising failures in many Unix utilities. Since then fuzzing has evolved into several approaches, including mutation-based fuzzing, generation-based fuzzing, and coverage-guided fuzzing. Mutation-based fuzzing starts from a corpus of existing inputs and mutates them, often guided by heuristics. Generation-based fuzzing builds inputs according to a formal specification or grammar. Coverage-guided fuzzing uses feedback from the program, such as code coverage, to steer input generation toward unexplored paths.

Typical fuzzing workflows involve creating a harness that exercises the target software, supplying a seed corpus, running a fuzzing loop that mutates or generates inputs, and monitoring for crashes or hangs. Reproducing crashes and triaging results are essential, often aided by sanitizers that detect memory errors, use-after-free conditions, and other issues.

Prominent fuzzing tools include AFL (American Fuzzy Lop) and libFuzzer, both of which support coverage-guided fuzzing, as well as other projects like honggfuzz. Fuzzing is used on user-space applications, kernels, browsers, network stacks, and APIs, and is commonly complemented by static analysis, formal verification, and other testing methods. Ethical and legal considerations apply when fuzzing third-party systems; authorization and responsible disclosure are important.