Fuzzing Code with AFL

Most programs are only ever used in fairly stereotyped ways on stereotyped input and will often crash in the presence of unexpected input. Test suites designed by humans, assuming there even is a test suite, are only as good as the people creating them and often only exercise the common code paths. This problem is where fuzzing comes in, the creation of input that exercises as many different code paths as possible in order to show up problems in the code. Until recently fuzzing has been a complex and tedious process, but with the appearance of instrumentation-guided fuzzers like AFL the task has become much easier. This article looks at how you can apply AFL to your code.