Improving Cryptocurrency Crime Detection: CoinJoin Community Detection Approach

Anton Wahrstatter, Jorao Gomes, Sajjad Khan, Davor Svetinovic

Research output: Contribution to journalArticlepeer-review

8 Scopus citations

Abstract

The potential of Bitcoin for money laundering and terrorist financing represents a significant challenge in law enforcement. In recent years, the use of privacy-improving CoinJoin transactions has grown significantly and helped criminal actors obfuscate Bitcoin money flows. In this study, we use unsupervised machine learning to analyze the complete Bitcoin user graph in order to identify suspicious actors potentially involved in illegal activities. In contrast to the existing studies, we introduce a novel set of features that we use to identify potential criminal activity more accurately. Furthermore, we apply our clustering algorithm to a CoinJoin-adjusted variant of the Bitcoin user graph, which enables us to analyze the network at a more detailed, user-centric level while still offering opportunities to address advanced privacy-enhancing techniques at a later stage. By comparing the results with our ground truth data set, we find that our improved clustering method is able to capture significantly more illicit activity within the most suspicious clusters. Finally, we find that users associated with illegal activities commonly have significant short paths to CoinJoin wallets and show tendencies toward outlier behavior. Our results have potential contributions to anti-money laundering efforts and combating the financing of terrorism and other illegal activities.

Original languageBritish English
Pages (from-to)1-11
Number of pages11
JournalIEEE Transactions on Dependable and Secure Computing
DOIs
StateAccepted/In press - 2023

Keywords

  • Bitcoin
  • Bitcoin
  • Blockchains
  • CoinJoins
  • Computer hacking
  • crime detection
  • cryptocurrency
  • Economics
  • Privacy
  • Support vector machines
  • unsupervised learning
  • Unsupervised learning

Fingerprint

Dive into the research topics of 'Improving Cryptocurrency Crime Detection: CoinJoin Community Detection Approach'. Together they form a unique fingerprint.

Cite this