ammaraskar/neuzz

neural network assisted fuzzer

NEUZZ: a neural-network-assisted fuzzer (S&P'19)

See IEEE S&P(Oakland)'19 slides and paper NEUZZ: Efficient Fuzzing with Neural Program Smoothing for details.

Prerequisite

Tested on a machine with Nvidia 1080Ti, Ubuntu 16.04/18.04, Tensorflow 1.8.0 and Keras 2.2.3.
We recommend running NEUZZ on a machine with a Nvidia 1080Ti or higher for efficient NN training.

• Python 2.7
• Tensorflow
• Keras

Build

    gcc -O3 -funroll-loops ./neuzz.c -o neuzz

Usage

We use a sample program readelf as an example.
Open a terminal, start nn module

    #python nn.py [program [arguments]]
    python nn.py ./readelf -a

open another terminal, start neuzz module.

    #./neuzz -i in_dir -o out_dir -l mutation_len [program path [arguments]] @@
    ./neuzz -i neuzz_in -o seeds -l 7506 ./readelf -a @@  

If you want to try NEUZZ on a new program,

1. Compile the new program from source code using afl-gcc.
2. Collect the training data by running AFL on the binary for a while(about an hour), then copy the queue folder to neuzz_in.
3. Follow the above two steps to start NN module and NEUZZ module.

Running with ASAN

If your binary is compiled with ASAN instrumentation, do the following to run it properly.

Pass --enable-asan to nn.py:

    python nn.py --enable-asan ./readelf -a

And pass -m none to ./nuezz as you would to afl:

    ./neuzz -m none -i neuzz_in -o seeds -l 7506 ./readelf -a @@

Sample programs

Try 10 real-world programs on NEUZZ. Check setup details at programs/[program names]/README.

Contact

Feel free to send me email about Neuzz. dongdong at cs.columbia.edu