Who is Gambling? Finding Cryptocurrency Gamblers Using Multi-modal Retrieval Methods

TL;DR

This paper introduces ETHGamDet, a tool that detects gambling-related smart contracts and addresses on blockchain using multi-modal retrieval and a novel LightGBM model, supported by a large-scale dataset.

Contribution

The paper presents a new detection tool, a novel LightGBM model with memory, and releases a large-scale dataset for identifying cryptocurrency gambling behaviors.

Findings

ETHGamDet achieves high classification F1-scores of 0.72 and 0.89.

The LightGBM model effectively learns from its misclassifications.

The large-scale dataset supports future research in blockchain gambling detection.

Abstract

With the popularity of cryptocurrencies and the remarkable development of blockchain technology, decentralized applications emerged as a revolutionary force for the Internet. Meanwhile, decentralized applications have also attracted intense attention from the online gambling community, with more and more decentralized gambling platforms created through the help of smart contracts. Compared with conventional gambling platforms, decentralized gambling have transparent rules and a low participation threshold, attracting a substantial number of gamblers. In order to discover gambling behaviors and identify the contracts and addresses involved in gambling, we propose a tool termed ETHGamDet. The tool is able to automatically detect the smart contracts and addresses involved in gambling by scrutinizing the smart contract code and address transaction records. Interestingly, we present a novel LightGBM model with memory components, which possesses the ability to learn from its own misclassifications. As a side contribution, we construct and release a large-scale gambling dataset at https://github.com/AwesomeHuang/Bitcoin-Gambling-Dataset to facilitate future research in this field. Empirically, ETHGamDet achieves a F1-score of 0.72 and 0.89 in address classification and contract classification respectively, and offers novel and interesting insights.