CodeAlchemist

CodeAlchemist is a JavaScript engine fuzzer that improves classic grammar-based JS engine fuzzers by a novel test case generation algorithm, called semantics-aware assembly. The details of the algorithm is in our paper, "CodeAlchemist: Semantics-Aware Code Generation to Find Vulnerabilities in JavaScript Engines", which appeared in NDSS 2019. This is a stable version of CodeAlchemist and it currently supports ChakraCore, V8, SpiderMonkey, and JavaScriptCore.

Installation

CodeAlchemist currently works on only Linux and we tested on Ubuntu 18.04.

1. Install dependencies (build-essential, nodejs, npm, esprima, dotnet)

$ sudo apt update
$ sudo apt install build-essential
$ sudo apt install nodejs npm
$ npm i [email protected]

Installation for dotnet depends on OS version, so please refer this link.

2. Clone and build CodeAlchemist

$ git clone https://github.com/SoftSec-KAIST/CodeAlchemist
$ cd CodeAlchemist
$ make

Usage

1. Prepare to start

Prepare JS seed files, a configuration file, and the requirements in the configuration. Please refer conf/README.md for writing the configuration file.

2. Preprocess JS seed files

$ dotnet bin/Main.dll rewrite <conf ABSPATH>
$ dotnet bin/Main.dll instrument <conf ABSPATH>

3. Run fuzzing process

$ dotnet bin/Main.dll fuzz <conf ABSPATH>

There are four optional parameters for our JS code generation algorithm.

• iMax (default: 8): The maximum number of iterations of the generation algorithm.
• pBlk (default: 16): The probability of reinventing block statements.
• iBlk (default: 3): The maximum number of iteration for generating a block statement.
• dMax (default: 3): The maximum nesting level for a reassembling block statement.

You can specify parameters with following commands.

$ dotnet bin/Main.dll fuzz <conf ABSPATH> --iMax 8 --pBlk 16 --iBlk 3 --dMax 3

CVEs (Credits)

If you find bugs and get CVEs by running CodeAlchemist, please let us know by sending a PR for ./docs/CVE.md.

• JavaScriptCore: CVE-2018-4464, CVE-2018-4437, CVE-2018-4378, CVE-2018-4372

Authors

This research project has been conducted by SoftSec Lab at KAIST.

• HyungSeok Han
• DongHyeon Oh
• Sang Kil Cha

Citation

If you plan to use CodeAlchemist in your own research. Please consider citing our paper:

@INPROCEEDINGS{han:ndss:2019,
  author = {HyungSeok Han and DongHyeon Oh and Sang Kil Cha},
  title = {{CodeAlchemist}: Semantics-Aware Code Generation to Find Vulnerabilities in JavaScript Engines},
  booktitle = ndss,
  year = 2019
}