sybil-scorer

Sybil scorer is a python package that provides useful classes and methods to analyze the behavior of addresses.

Help Out

Consider donating on Gitcoin https://explorer.gitcoin.co/#/round/10/0x8de918f0163b2021839a8d84954dd7e8e151326d/0x8de918f0163b2021839a8d84954dd7e8e151326d-66

Installation

• Python >= 3.9
• pip install sybil-scorer

What is inside?

The package has two main sub-packages.

• sbdata is a package to easily retrieve a large amount of data from the Flipside API.
• sblegos a package to perform on-chain transactions analysis to detect potential Sybil behavior.
• sbutils is a package that makes it easy to load the data extracted with sbdata and use it in sblegos.

More details on the packages and examples are provided below.

sbdata

An example script to retrieve data from the flipside API is provided in the script folder: script/demo_extract_eth_txs_oss.py

It could also be used to retrieve transactional data from an address with some simple calls:

import os
from sbdata.FlipsideApi import FlipsideApi

api_key = os.environ['FLIPSIDE_API_KEY']
flipside_api = FlipsideApi(api_key, max_address=1000)
address = ['0x06cd8288dc001024ce0a1cf39caaedc0e2db9c82']
tx_add_eth = flipside_api.get_transactions(address, chain='ethereum')

It walks you through the process of retrieving data from the flipside API and saving it in a folder.

To use this package you will need an API key from flipside that you can get here: https://sdk.flipsidecrypto.xyz/shroomdk/apikeys

Useful example script are provided in the script folder.

sblegos and sbutils

sblegos provides the following analysis of legos:

• has_same_seed : true if the address has the same seed as any other address in the grants contributors
• has_same_seed_naive : true if the address has the same seed as any other address in the grants contributors with a naive approach: address of the from_address of the first transaction.
• has_suspicious_seed_behavior : true if has_same_seed is different from has_same_seed_naive. It means the user performed some actions before funding his wallet.
• has_interacted_with_other_contributor : true if the user has interacted with any other contributor to the grant
• has_less_than_n_transactions : true if the user has less than n transactions.
• has_transaction_similitude : true if the user has a transaction history that is similar to any other contributor to the grant.
• has_transaction_similitude_opti : an optimized version of has_transaction_similitude, when used across multiple addresses.

A jupyter notebook using both packages is available as a jupyter notebook here https://github.com/poupou-web3/grant-exploration/blob/main/gr-climate-exploration.ipynb

The following snippet of code will check if any address has the same seed as any other contributor to the climate grant

import os
from pathlib import Path
import numpy as np
import pandas as pd
from sbutils import LoadData
from sblegos.TransactionAnalyser import TransactionAnalyser as txa

# Set path to data folder
PATH_TO_EXPORT = 'path to where the data was extracted'
CHAIN = 'the name of the chain you want to analyse for example "ethereum"'

# Load the votes data
array_unique_address = df_votes['voter'].unique() # array of unique voters, here df_votes contains all the votes made on a grant 

# Be sure that the address are in lower case
array_unique_address = np.char.lower(array_unique_address.astype(str))
print(f'Number of unique voter: {len(array_unique_address)}')

# Load the transactions of the addresses using the sbutils package
data_loader = LoadData.LoadData(PATH_TO_EXPORT)
df_tx = data_loader.create_df_tx(CHAIN, array_unique_address)

# Initialise the TransactionAnalyser class
tx_analyser = txa(df_tx, array_address=array_unique_address)

# Compute some predetermined features, it can takes some time especially on large datasets
df_matching_address = tx_analyser.get_df_features()
df_matching_address.head(2)

# For individual computation of the features:
df_matching_address = pd.DataFrame(array_unique_address, columns=["address"])
df_matching_address['seed_same_naive'] = df_matching_address.loc[:, 'address'].apply(lambda x : tx_analyser.has_same_seed_naive(x))

Documentation

The documentation of the package is available at https://sybil-scorer.readthedocs.io/en/latest/py-modindex.html. For a local version of the documentation, you can build it using sphinx. with the following commands:

cd docs
sphinx-apidoc -o ./source ../sbscorer
make html

Then open the file docs/build/html/index.html in your browser. The local version of the documentation is prettier than the one hosted on readthedocs.

Example Data

Some data for easier use of the package in the context of Gitcoin grants are made available on Ocean market.

Gitcoin Citizen Round data

Transaction data

You can load the data directly into the df_tx variable. https://huggingface.co/datasets/Poupou/Gitcoin-Citizen-Round/blob/main/tx_citizen_round.parquet

These are all the transactions performed by users who contributed to the Citizen round on grant as of 30th of June 2023.

Example of vote data

Example query to extract the vote data of the citizen round of June 2023 from the Flipside API: https://flipsidecrypto.xyz/poupou/q/j3E9SEfMLkxG/citizen-round-votes

You could also use the data available on hugging face: https://huggingface.co/datasets/Poupou/Gitcoin-Citizen-Round/blob/main/citizen-votes.csv

Future works

Future works include:

• Adding more transactional analysis lego.
• Adding temporal features to a clustering algorithm as researched in the first hackathon submission.
• Improving seed legos to output cluster of addresses instead of a boolean.
• See issues at www.github.com/poupou-web3/sybil-scorer/issues