A Flash(bot) in the Pan

A collection of tools to measure and analyze frontrunning attacks on private pools such as Flashbots. Our paper can be found here and our data is available for download here.

Quick Start

A container with all the dependencies can be found here.

To run the container, please install docker and run:

docker pull christoftorres/a-flashbot-in-the-pan && docker run -m 16g --memory-swap="24g" -p 8888:8888 -it christoftorres/a-flashbot-in-the-pan

Afterwards, start an instance of MongoDB inside the container:

mkdir -p /data/db && mongod --fork --logpath /var/log/mongod.log

Import the flashbots blocks into MongoDB by running inside the container the following commands:

cd /root/data-collection/flashbots
python3 import_flashbots_data.py

To run the MEV measurement scripts, simply run inside the container the following commands:

# Run the sandwich measurement script
cd /root/data-collection/mev/sandwiches
python3 sandwiches.py <BLOCK_RANGE_START>:<BLOCK_RANGE_END> # For exmaple: python3 sandwiches.py 10892526:10892526

# Run the arbitrage measurement script
cd /root/data-collection/mev/arbitrage
python3 arbitrage.py <BLOCK_RANGE_START>:<BLOCK_RANGE_END> # For exmaple: python3 arbitrage.py 11706655:11706655

# Run the liquidation measurement script
cd /root/data-collection/mev/liquidation
python3 liquidation.py <BLOCK_RANGE_START>:<BLOCK_RANGE_END> # For exmaple: python3 liquidation.py 11181773:11181773

To collect pending transactions, simply run inside the container the following commands:

cd /root/data-collection/pending-transactions
node observer.js

To run the analysis, please launch the Jupyter notebook server inside the container using the following commands and then open up http://localhost:8888 on your browser:

cd /root/analysis
jupyter notebook --port=8888 --no-browser --ip=0.0.0.0 --allow-root --NotebookApp.token='' --NotebookApp.password=''

Custom Docker image build

docker build -t a-flashbot-in-the-pan .
docker run -m 16g --memory-swap="24g" -p 8888:8888 -it a-flashbot-in-the-pan:latest

Installation Instructions

1. Install MongoDB

MacOS

brew tap mongodb/brew
brew install mongodb-community@4.4

For other operating systems follow the installation instructions on mongodb.com.

2. Install Python dependencies

python3 -m pip install -r requirements.txt

3. Install Node.js

MacOS

brew install node

For other operating systems follow the installation instructions on nodejs.org.

Data Collection

Launch MongoDB

mongod

Download and import Flashbots data

cd data-collection/flashbots
python3 import_flashbots_data.py

Measure MEV extraction

⚠️ !! To measure past MEV extraction you will need a connection to a fully synched Ethereum archive node and change PROVIDER in data-collection/mev/utils/settings.py accordingly. !!

cd data-collection/mev/sandwiches
python3 sandwiches.py  <BLOCK_RANGE_START>:<BLOCK_RANGE_END> 
cd data-collection/mev/arbitrage
python3 arbitrage.py   <BLOCK_RANGE_START>:<BLOCK_RANGE_END> 
cd data-collection/mev/liquidation
python3 liquidation.py <BLOCK_RANGE_START>:<BLOCK_RANGE_END> 

Collect pending transactions

⚠️ !! To collect pending transactions you will need a connection to an Ethereum node and set web3 in data-collection/pending-transactions/observer.js accordingly. !!

cd data-collection/pending-transactions
npm install
node observer.js

Analysis

You can either run the data collection scripts or download our data from Google drive:

cd data-collection
./download_and_import_data.sh

The bulk of the analysis was done in Jupyter notebooks, which can be opened by running:

cd analysis
jupyter notebook

and selecting the notebook of choice.

Attribution

If using this repository for research, please cite as

@inproceedings{
  aflashbotinthepan, 
  address={Nice, France}, 
  title={A Flash(bot) in the Pan: Measuring Maximal Extractable Value in Private Pools}, 
  ISBN={978-1-4503-9259-4}, 
  DOI={10.1145/3517745.3561448}, 
  booktitle={Proceedings of the 22nd ACM Internet Measurement Conference (IMC ’22)}, 
  publisher={Association for Computing Machinery}, 
  author={Weintraub, Ben and Ferreira Torres, Christof and Nita-Rotaru, Cristina and State, Radu}, 
  year={2022} 
}