JoinDetect: A Data-Led Machine Learning Approach to Detection of Coinjoin Transactions
Open Access
- 4 September 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 5-13
Abstract
CoinJoin is a privacy-enhancing technique used in cryptocurrency transactions, designed to obfuscate the ownership of funds by amalgamating multiple inputs, thereby complicating the task of linking identities to addresses. While traditional methods for detecting such transactions have relied on rule-based and heuristic approaches, these often fall short in addressing the complexities of advanced CoinJoin techniques. Recent advancements have shown that machine learning methods significantly outperform these traditional approaches in identifying these types of transactions. This paper introduces a robust data engineering and machine learning framework to effectively distinguish CoinJoin transactions from standard ones. Our approach tackles class imbalance, and accounts for varying frequency distributions of target class transactions. We detail a two-part process: initially demonstrating the approach’s utility in identifying classic Chaumian CoinJoin transactions, then reapplying the methodology for model retraining to adapt to concept drift, enabling the detection of newer, more complex CoinJoin variants. Leveraging ground truth labelled data and proprietary blockchain analytics, our model adeptly navigates the evolving challenges of privacy technologies. Beyond mere identification, this process enhances blockchain forensic applications, improving blockchain address clustering and the efficiency of attribution tracing. This not only streamlines forensic investigations, but also supports regulatory compliance initiatives, ensuring that the application of blockchain technology aligns with broader societal values without compromising user privacy. By integrating blockchain technology in this manner, we contribute to a framework where digital finance meets social good, providing tools that ensure both innovation and integrity in blockchain applications.Keywords
Funding Information
- Science Foundation Ireland (18/CRT/6183)
This publication has 14 references indexed in Scilit:
- The next phase of identifying illicit activity in BitcoinInternational Journal of Network Management, 2024
- Identifying illicit transactions in Bitcoin Tumbler services using Supervised Machine Learning Algorithms2023
- Cryptocurrency Concepts, Technology, and Applications2023
- Improving Cryptocurrency Crime Detection: CoinJoin Community Detection ApproachIEEE Transactions on Dependable and Secure Computing, 2023
- Adoption and Actual Privacy of Decentralized CoinJoin Implementations in Bitcoin2022
- The Unique Dressing of Transactions: Wasabi CoinJoin Transaction Detection2022
- CoinJoin in the Wild2021
- Unnecessary Input Heuristics and PayJoin Transactions2021
- LSTM-TC: Bitcoin coin mixing detection method with a high recallApplied Intelligence, 2021
- A fistful of BitcoinsCommunications of the ACM, 2016