Multilayer networks receive increasing attention in the last few years, due to their flexibility in modeling interconnected systems. There also exist several data repositories with multilayer graphs, but neither of them has publicly available benchmark data on multilayer graphs with ground truth for anomaly detection. In this task, the goal is to detect anomalies in a dynamic network.

Topological Analysis for Event Detection Dataset

We provide a graph dataset that was extracted from the Ethereum blockchain. We only include tokens reported by the EtherScan.io online explorer to have more than $100M in market value. Eventually, the data set contains 6 tokens, and on average, each token has a history of 297 days (minimum and maximum of 151 and 576 days, respectively). As ground truth, we adopt and curate Blockchain events from Wikipedia, which lists and explains major events since 2008. In total, there are 72 events that have shaped blockchain networks — some of them in adverse (see the supplementary material for the complete list). However, token networks cannot detect events before 2015 because Ethereum and its tokens did exist before then. Hence, our experiments focused on at most 32 (out of the 72) token transaction events.

                

@InProceedings{anomalyMultilayerOfori, author="Ofori-Boateng, D. and Dominguez, I. Segovia and Akcora, C. and Kantarcioglu, M. and Gel, Y. R.", editor="Oliver, Nuria and Perez-Cruz, Fernando and Kramer, Stefan and Read, Jesse and Lozano, Jose A.", title="Topological Anomaly Detection in Dynamic Multilayer Blockchain Networks", booktitle="Machine Learning and Knowledge Discovery in Databases. Research Track", year="2021", publisher="Springer International Publishing"}

Full Dataset 

Data Set Characteristics: graph files
Task: Classification, Clustering

Data start date (UTC): 2015
Data end date (UTC): 2018
Number of token networks: 6

Prediction task: Predict events given in the file.
Networks are in the format of fromAddress\ttoAddress\tunixtime\tAmount.

Each network file is less than 1MB in size.

Bytom network

Cybermiles network

Decentraland network

Tierion network

Vechain network

ZRX network

Blockchain Events (anomaly sources) can be found here: events file

Extended Dataset 

We are sharing an extended data set that is not analyzed in the referenced article. The dataset contains 1701 token networks

Download Extended Dataset

Cite Our Dataset:

	@inproceedings{chartalistNeurips2022,
  author    = {Kiarash Shamsi and Yulia R. Gel and  Murat Kantarcioglu and Cuneyt G. Akcora},
  title     = {Chartalist: Labeled Graph Datasets for UTXO and Account-based Blockchains},
  booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference
               on Neural Information Processing Systems 2022, NeurIPS 2022, November 29-December
               1, 2022, New Orleans, LA, USA},
  pages     = {1--14},
  year      = {2022},
  url       = {https://openreview.net/pdf?id=10iA3OowAV3}
  }

Baseline

topological anomaly detection (TAD) framework for dynamic multilayer networks. TAD yields a highly competitive performance in detecting anomalous events on Ethereum and Ripple blockchains.

Baseline Reference Paper

	@InProceedings{anomalyMultilayerOfori,
author="Ofori-Boateng, D. and Dominguez, I. Segovia and Akcora, C. and Kantarcioglu, M. and Gel, Y. R.",
editor="Oliver, Nuria and Perez-Cruz, Fernando and Kramer, Stefan and Read, Jesse and Lozano, Jose A.",
title="Topological Anomaly Detection in Dynamic Multilayer Blockchain Networks",
booktitle="Machine Learning and Knowledge Discovery in Databases. Research Track",
year="2021",
publisher="Springer International Publishing"}