Experiments in Graph Deep Learning for Cryptocurrency Forensics

Applying AI techniques to identify and combat money laundering.

Access dataset

Read paper

About

 

Dataset

To motivate and enable the development of new techniques for detection of illicit cryptocurrency transactions, Elliptic has released one of the world’s largest sets of labeled transaction data publicly available in any cryptocurrency. The Elliptic Data Set consists of over 200,000 bitcoin transactions with a total value of $6 billion. Transactions identified by Elliptic as having been made by criminal actors have been labeled to allow the development and testing of new predictive techniques.

Download dataset



Highlights

The Elliptic Data Set consists of over 200,000 bitcoin transactions with a total value of $6 billion. This anonymized data set is a transaction graph collected from the Bitcoin blockchain. 

The dataset is one of the world’s largest sets of labeled transaction data publicly available in any cryptocurrency.  

The dataset (CSV file format) can be downloaded under a CC BY-NC-ND 4.0 license and contains maps Bitcoin transactions to real entities belonging to licit categories (exchanges, wallet providers, miners, licit services, etc.) versus illicit ones (scams, malware, terrorist organizations, ransomware, Ponzi schemes, etc.).

The goal of the data set is to further the development and testing of new predictive techniques to combat money laundering. Users can use the dataset to classify the illicit and licit nodes in the transaction graph. 

Research

In our paper entitled “Anti-Money Laundering in Bitcoin: Experimenting with Graph Convolutional Networks for Financial Forensics,” we share our motivation, approach, and progress in developing and applying new machine learning methods such as Graph Convolutional Networks (GCNs) and Random Forest (RF) to Anti-money laundering forensic investigation in Bitcoin. We introduce and describe the Elliptic Data Set, a graph network of over 200,000 Bitcoin transactions with handcrafted features. We hope to a) inform and inspire others to work on this societally important challenge, and b) invite technical and domain feedback in support of our ongoing inquiry.

Access paper



Research team

 
researcher headshot

Jie Chen (Primary investigator)

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Charles E. Leiserson (Primary investigator)

MIT-IBM Watson AI Lab

MIT CSAIL

researcher headshot

Toyotaro Suzumura (Primary investigator)

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Giacomo Domeniconi

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Tim Kaler

MIT-IBM Watson AI Lab

MIT CSAIL

researcher headshot

Hiroki Kanezashi

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Georgios Kollias

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Tengfei Ma

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Aldo Pareja

MIT-IBM Watson AI Lab

IBM Research

researcher headshot

Tao B. Schardl

MIT-IBM Watson AI Lab

MIT CSAIL

researcher headshot

Mark Weber

MIT-IBM Watson AI Lab

IBM Research

Explore our research

The scale of the research effort between MIT and IBM is unprecedented in industry and academia. We’ve already made exciting progress, funding and launching more than 50 projects selected from 186 high-quality proposals received from nearly every MIT department. We believe these initiatives could drive some of the most significant advancements in AI.

Explore research