OSS projects have different build and test systems. We can’t expect them all to implement and maintain fuzz targets or integrate them with OSS-Fuzz in the same way. However, we do have recommendations.
This page documents several features (starting from the easiest) that will make automated fuzzing simple and efficient, and will help you catch regressions early in the development cycle. This simple example covers most of the items.
- Fuzz Target
- Build support
- Seed Corpus
- Regression Testing
- Not a project member?
Every fuzz target:
- Is maintained by code owners in their RCS (Git, SVN, etc).
- Is built with the rest of the tests - no bit rot!
- Has a seed corpus with good code coverage.
- Has a dictionary, if applicable.
- Is continuously tested on the seed corpus with ASan/UBSan/MSan.
- Is fast and has no OOMs.
The code of the fuzz target(s) should be part of the project’s source code repository. All fuzz targets should be easily discoverable (reside in the same directory, follow the same naming pattern, etc.).
This makes it easy to maintain the fuzzers and minimizes breakages that can arise as source code changes over time.
Make sure to fuzz the target locally for a small period of time to ensure that it does not crash, hang, or run out of memory instantly. If you’re having trouble, read about what makes a good fuzz target.
The interface between the fuzz target and the fuzzing engines is C, so you can use either C or C++ to implement the fuzz target.
Many different build systems exist in the open-source world. The less OSS-Fuzz knows about them, the better it can scale.
An ideal build integration for OSS-Fuzz looks like this:
- For every fuzz target
fooin the project, there is a build rule that builds
foo_fuzzer, a binary that:
- Contains the fuzzing entry point.
- Contains (
LLVMFuzzerTestOneInput) and all the code it depends on
- Uses the
$LIB_FUZZING_ENGINE(env var provided by OSS-Fuzz environment).
- Since the build system supports changing the compiler and passing extra compiler flags, the build command for
foo_fuzzerlooks similar to this:
# Assume the following env vars are set: # CC, CXX, CFLAGS, CXXFLAGS, LIB_FUZZING_ENGINE $ make_or_whatever_other_command foo_fuzzer
This minimizes OSS-Fuzz-specific configuration, making your fuzzing more robust.
There is no point in hardcoding the exact compiler flags in the build system because they a) may change and b) depend on the fuzzing engine and sanitizer being used.
The seed corpus is a set of test inputs, stored as individual files, provided to the fuzz target as a starting point (to “seed” the mutations). The quality of the seed corpus has a huge impact on fuzzing efficiency; the higher the quality, the easier it is for the fuzzer to discover new code paths. The ideal corpus is a minimal set of inputs that provides maximal code coverage.
For better OSS-Fuzz integration, the seed corpus should be available in revision control (it can be the same as or different from the source code). It should be regularly extended with the inputs that (used to) trigger bugs and/or touch new parts of the code.
For some input types, a simple dictionary of tokens used by the input language can have a dramatic impact on fuzzing efficiency. For example, when fuzzing an XML parser, a dictionary of XML tokens is helpful. AFL has a collection of dictionaries for popular data formats. Ideally, a dictionary should be maintained alongside the fuzz target, and it must use correct syntax.
For a fuzz target to be useful, it must have good coverage in the code that it is testing. You can view the coverage for your fuzz targets by looking at the fuzzer stats dashboard on ClusterFuzz, as well as coverage reports.
To generate an aggregated code coverage report for your project, please see the code coverage page.
Coverage can often be improved by adding dictionaries, more inputs for seed corpora, and fixing timeouts/out-of-memory bugs in your targets.
Fuzz targets should be regularly tested (not necessarily fuzzed!) as a part of the project’s regression testing process. One way to do so is to link the fuzz target with a simple standalone driver (example) that runs the provided inputs, then use this driver with the seed corpus created in previous step. We recommend you use sanitizers during regression testing.
Fuzz targets should perform well, because high memory usage and/or slow execution speed can slow the down the growth of coverage and finding of new bugs. ClusterFuzz provides a performance analyzer for each fuzz target that shows problems that are impacting performance.
If you are a member of the project you want to fuzz, most of the steps above are simple. However in some cases, someone outside the project team may want to fuzz the code, and the project maintainers are not interested in helping.
In such cases, we can host the fuzz targets, dictionaries, etc. in OSS-Fuzz’s repository and mention them in the Dockerfile. It’s not ideal, because the fuzz targets will not be continuously tested, so may quickly bitrot.
If you are not a project maintainer, we may not be able to CC you to security bugs found by OSS-Fuzz.