Advanced Fuzzing Library - Slot your own fuzzers together and extend their features using Rust.

LibAFL is written and maintained by

• Andrea Fioraldi [email protected]
• Dominik Maier [email protected]
• s1341 [email protected]
• Dongjia Zhang [email protected]
• Addison Crump [email protected]

LibAFL gives you many of the benefits of an off-the-shelf fuzzer, while being completely customizable. Some highlight features currently include:

• fast: We do everything we can at compile time, keeping runtime overhead minimal. Users reach 120k execs/sec in frida-mode on a phone (using all cores).
• scalable: Low Level Message Passing, LLMP for short, allows LibAFL to scale almost linearly over cores, and via TCP to multiple machines.
• adaptable: You can replace each part of LibAFL. For example, BytesInput is just one potential form input: feel free to add an AST-based input for structured fuzzing, and more.
• multi platform: LibAFL was confirmed to work on Windows, MacOS, Linux, and Android on x86_64 and aarch64. LibAFL can be built in no_std mode to inject LibAFL into obscure targets like embedded devices and hypervisors.
• bring your own target: We support binary-only modes, like Frida-Mode, as well as multiple compilation passes for sourced-based instrumentation. Of course it's easy to add custom instrumentation backends.

LibAFL is a collection of reusable pieces of fuzzers, written in Rust. It is fast, multi-platform, no_std compatible, and scales over cores and machines.

It offers a main crate that provide building blocks for custom fuzzers, libafl, a library containing common code that can be used for targets instrumentation, libafl_targets, and a library providing facilities to wrap compilers, libafl_cc.

LibAFL offers integrations with popular instrumentation frameworks. At the moment, the supported backends are:

• SanitizerCoverage, in libafl_targets
• Frida, in libafl_frida
• QEMU user-mode and system mode, including hooks for emulation, in libafl_qemu
• TinyInst, in libafl_tinyinst by elbiazo

1. Install the Dependecies

• The Rust development language.
  We highly recommend not to use e.g. your Linux distribition package as this is likely outdated. So rather install Rust directly, instructions can be found here.

• LLVM tools
  The LLVM tools (including clang, clang++) are needed (newer than LLVM 11.0.0 up to LLVM 17.0.0)

• Cargo-make
  We use cargo-make to build the fuzzers in fuzzers/ directory. You can install it with

cargo install cargo-make

2. Clone the LibAFL repository with

git clone https://github.com/AFLplusplus/LibAFL

3. Build the library using

cargo build --release

4. Build the API documentation with

cargo doc

5. Browse the LibAFL book (WIP!) with (requires mdbook)

cd docs && mdbook serve

We collect all example fuzzers in ./fuzzers. Be sure to read their documentation (and source), this is the natural way to get started!

You can run each example fuzzer with

cargo make run

as long as the fuzzer directory has Makefile.toml file.

The best-tested fuzzer is ./fuzzers/libfuzzer_libpng, a multicore libfuzzer-like fuzzer using LibAFL for a libpng harness.

• Installation guide

• Online API documentation

• The LibAFL book (WIP) online or in the repo

• Our research paper

• Our RC3 talk explaining the core concepts

• Our Fuzzcon Europe talk with a (a bit but not so much outdated) step-by-step discussion on how to build some example fuzzers

• The Fuzzing101 solutions & series of blog posts by epi

• Blogpost on binary-only fuzzing lib libaf_qemu, Hacking TMNF - Fuzzing the game server, by RickdeJager.

• A LibAFL Introductory Workshop, by Jordan Whitehead

For bugs, feel free to open issues or contact us directly. Thank you for your support. <3

Even though we will gladly assist you in finishing up your PR, try to

• keep all the crates compiling with stable rust (hide the eventual non-stable code under cfgs)
• run cargo fmt on your code before pushing
• check the output of cargo clippy --all or ./clippy.sh
• run cargo build --no-default-features to check for no_std compatibility (and possibly add #[cfg(feature = "std")]) to hide parts of your code.

Some of the parts in this list may be hard, don't be afraid to open a PR if you cannot fix them by yourself, so we can help.

If you use LibAFL for your academic work, please cite the following paper:

@inproceedings{libafl,
 author       = {Andrea Fioraldi and Dominik Maier and Dongjia Zhang and Davide Balzarotti},
 title        = {{LibAFL: A Framework to Build Modular and Reusable Fuzzers}},
 booktitle    = {Proceedings of the 29th ACM conference on Computer and communications security (CCS)},
 series       = {CCS '22},
 year         = {2022},
 month        = {November},
 location     = {Los Angeles, U.S.A.},
 publisher    = {ACM},
}