Awesome

MOCK

We open source the prototype of Mock here. Mock is a Linux kernel fuzzer that can learn the contextual dependencies in syscall sequences and then generate context-aware test cases. In this way, Mock improves the input quality and explore deeper space of kernels. More details can be found in our paper from NDSS'24.

Installation

As Mock is built upon Healer, please follow the instructions to prepare necessary toolchains. Image and kernel preparation can be found in this document.

Besides, the training component is written in Python and interacts with the fuzzing component via http. Therefore, Python packages should be installed.

> pip3 install numpy django 
# if cuda is available
> pip3 install torch torchvision torchaudio \
    --index-url https://download.pytorch.org/whl/cu113

Once all the required tools have been installed, Mock can be easily built using the following command. It may some take to prepare the Rust bindings for Libtorch, tch-rc, on which Mock depends.

> cargo build --release

You can find Mock and the patched Syzkaller binary (syz-bin) can be found in the target/release directory.

Usage

The first step is run Mock is to launch a django service. It serves to monitor requests for the parallel model training. Start it using the following command:

> cd $MOCK_ROOT/tools/model_manager
> python3 manage.py runserver

Suppose that everything is fine (instructions here), execute following command in a new terminal to start the fuzzing (-d specifies the path to disk image, -k specifies the path to kernel image and --ssh-key specifies the path to ssh key).

> healer -d stretch.img --ssh-key stretch.id_rsa -k bzImage

Citation

@inproceedings{
    author = {Jiacheng, Xu and Xuhong, Zhang and Shouling, Ji and Yuan, Tian and Binbin, Zhao and Qinying, Wang and Peng, Cheng and Jiming, Chen}, 
    title = {MOCK: Optimizing Kernel Fuzzing Mutation with Context-aware Dependency},
    booktitle = {31st Annual Network and Distributed System Security Symposium, {NDSS} 2024, San Diego, California, USA, February 26 - March 1, 2024}, 
    publisher = {The Internet Society},
    year = {2024}, 
}