This repository contains the official code for the paper "GAMLNet: a graph based framework for the detection of money launderings." accepted at IEEE Swiss Data Science Conference 2024 (SDS24).
You will need to install the requred dependencies before running the code. You can create the conda environment with conda env create -f GAMLNet_env.yml
. After you'll need to activate the environment with conda activate GAMLNet
.
A tutorial jupyter notebook on how to run the code is provided in the main directory. This shows how to load datasets and run the model.
The official datasets used in the paper can be downloaded here and should be placed in the datasets/
folder. The datasets datasets/8K_5
and datasets/16K_5
are already provided by default in this repository.
The table below shows additional information regarding these datasets.
Dataset 8K_5 visualzied below (8,000 nodes with 5% anomaly). Nodes in red are anomalous, nodes in gray are benign.
We propose the GAMLNet (Graph Anti-Money Laundering Network) architecture, leveraging the strengths of two popular GNN variants: Graph Isomorphism Networks (GIN) [Xu et al., 2019] and GraphSAGE [Hamilton et al., 2017]. GIN performs exceptionally well at learning isomorphic graph substructures and GraphSAGE performs exceptionally well in environments with rich statistical node feature information.