Coder Social home page Coder Social logo

namqdam / ityfuzz Goto Github PK

View Code? Open in Web Editor NEW

This project forked from fuzzland/ityfuzz

0.0 0.0 0.0 1.59 MB

Blazing Fast Hybrid Fuzzer for Smart Contracts

License: MIT License

Shell 0.02% Python 5.99% Rust 58.48% HTML 0.05% Dockerfile 0.19% Solidity 35.26%

ityfuzz's Introduction

ItyFuzz 🍦

Fast hybrid fuzzer for EVM & MoveVM (WIP) smart contracts.

You can generate exploits instantly by just providing the contract address:

中文版README / Research Paper / Development Info

Run ItyFuzz with UI

Install Docker and run docker image suitable for your system architecture:

docker pull fuzzland/ityfuzz:stable
docker run -p 8000:8000 fuzzland/ityfuzz:stable

Then, you can visit the interface at http://localhost:8000

Note: The container uses public ETH RPC, may time out / be slow

Statistics & Comparison

Time taken for finding vulnerabilities / generating exploits:

Project Name Vulnerability Mythril SMARTIAN Slither ItyFuzz
AES Business Logic Inf Unsupported No 4hrs
Carrot Arbitrary External Call 17s 11s Yes 1s
Olympus Access Control 36s Inf Yes 1s
MUMUG Price Manipulation Inf Unsupported No 18hrs
Omni Reentrancy Inf Unsupported Yes* 22hrs
Verilog CTF-2 Reentrancy Inf Unsupported Yes* 3s

* Slither only finds the reentrancy location, but not how to leverage reentrancy to trigger final buggy code. The output also contains significant amount of false positives.

Test Coverage:

Dataset SMARTIAN Echidna ItyFuzz
B1 97.1% 47.1% 99.2%
B2 86.2% 82.9% 95.4%
Tests Unsupported 52.9% 100%

* B1 and B2 contain 72 single-contract projects from SMARTIAN artifacts. Tests are the projects in tests directory. The coverage is calculated as (instruction covered) / (total instruction - dead code).

Development

Building

You need to have libssl-dev (OpenSSL) and libz3-dev (refer to Z3 Installation section for instruction) installed.

# download move dependencies
git submodule update --recursive --init
cd cli/
cargo build --release

You can enable certain debug gates in Cargo.toml

solc is needed for compiling smart contracts. You can use solc-select tool to manage the version of solc.

Run

Compile Smart Contracts:

cd ./tests/multi-contract/
# include the library from ./solidity_utils for example
solc *.sol -o . --bin --abi --overwrite --base-path ../../

Run Fuzzer:

# if cli binary exists
cd ./cli/
./cli -t '../tests/multi-contract/*'

Demo

Verilog CTF Challenge 2 tests/verilog-2/

Flashloan attack + Reentrancy. The target is to reach line 34 in Bounty.sol.

Exact Exploit:

0. Borrow k MATIC such that k > balance() / 10
1. depositMATIC() with k MATIC
2. redeem(k * 1e18) -- reentrancy contract --> getBounty()
3. Return k MATIC

Use fuzzer to detect the vulnerability and generate the exploit (takes 0 - 200s):

# build contracts in tests/verilog-2/
solc *.sol -o . --bin --abi --overwrite --base-path ../../
# run fuzzer
./cli -f -t "./tests/verilog-2/*"

-f flag enables automated flashloan, which hooks all ERC20 external calls and make any users to have infinite balance.

Fuzz a Project (Offline)

You can fuzz a project by providing a path to the project directory.

./cli -t '[DIR_PATH]/*'

ItyFuzz would attempt to deploy all artifacts in the directory to a blockchain with no other smart contracts.

Specifically, the project directory should contain a few [X].abi and [X].bin files. For example, to fuzz a contract named main.sol, you should ensure main.abi and main.bin exist in the project directory. The fuzzer will automatically detect the contracts in directory, the correlation between them (see tests/multi-contract), and fuzz them.

Optionally, if ItyFuzz fails to infer the correlation between contracts, you can add a [X].address, where [X] is the contract name, to specify the address of the contract.

Caveats:

  • Keep in mind that ItyFuzz is fuzzing on a clean blockchain, so you should ensure all related contracts (e.g., ERC20 token, Uniswap, etc.) are deployed to the blockchain before fuzzing.

  • You also need to overwrite all constructor(...) in the smart contract to to make it have no function argument. ItyFuzz assumes constructors have no argument.

Fuzz a Project (Online)

(Optional) Rebuild with flashloan_v2 (only supported in onchain) enabled to get better result.

sed -i 's/\"default = [\"/\"default = [flashloan_v2,\"/g' ./Cargo.toml
cd ./cli/
cargo build --release

(Optional, Must do if flashloan_v2 enabled) To effectively cache the costly RPC calls to blockchains, third-party APIs, and Etherscan, a proxy is built. To run the proxy:

pip3 install -r proxy/requirements.txt
python3 proxy/main.py

You can fuzz a project by providing an address, a block, and a chain type.

./cli -o -t [TARGET_ADDR] --onchain-block-number [BLOCK] -c [CHAIN_TYPE] 

Example: Fuzzing WETH contract (0xc02aaa39b223fe8d0a0e5c4f27ead9083c756cc2) on Ethereum mainnet at latest block.

./cli -o -t 0xc02aaa39b223fe8d0a0e5c4f27ead9083c756cc2 --onchain-block-number 0 -c ETH

Fuzzing with flashloan and oracles enabled:

./cli -o -t 0xc02aaa39b223fe8d0a0e5c4f27ead9083c756cc2 --onchain-block-number 0 -c ETH -f -i -p

ItyFuzz would pull the ABI of the contract from Etherscan and fuzz it. If ItyFuzz encounters an unknown slot in the memory, it would pull the slot from chain RPC. If ItyFuzz encounters calls to external unknown contract, it would pull the bytecode and ABI of that contract. If its ABI is not available, ItyFuzz would not send any transaction to that contract.

To use proxy, append --onchain-local-proxy-addr http://localhost:5003 to your CLI command.

Onchain Fetching

ItyFuzz attempts to fetch storage from blockchain nodes when SLOAD is encountered and the target is uninitialized. There are three ways of fetching:

  • OneByOne: fetch one slot at a time. This is the default mode. It is slow but never fails.
  • All: fetch all slots at once using custom API eth_getStorageAll on our nodes. This is the fastest mode, but it may fail if the contract is too large.
  • Dump: dump storage using debug API debug_storageRangeAt. This only works for ETH (for now) and fails most of the time.

Z3 Installation

macOS

git clone https://github.com/Z3Prover/z3 && cd z3
python scripts/mk_make.py --prefix=/usr/local
cd build && make -j64 && sudo make install

If the build command still fails for not finding z3.h, do export Z3_SYS_Z3_HEADER=/usr/local/include/z3.h

Ubuntu

apt install libz3-dev

Telemetry

ItyFuzz collects telemetry data to help us improve the fuzzer. The data is collected anonymously and is not used for any commercial purpose. You can disable telemetry by setting NO_TELEMETRY=1 in your environment variable.

Acknowledgement

This work was supported in part by NSF grants CCF-1900968, CCF1908870, and CNS1817122 and SKY Lab industrial sponsors and affiliates Astronomer, Google, IBM, Intel, Lacework, Microsoft, Mohamed Bin Zayed University of Artificial Intelligence, Nexla, Samsung SDS, Uber, and VMware. Any opinions, findings, conclusions, or recommendations in this repo do not necessarily reflect the position or the policy of the sponsors.

Grants:

Grants Description
Grants from Sui Foundation for building Move and chain-specfic support
Grants from Web3 Foundation for building Substrate pallets and Ink! support

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.