Ilmu Komputer & AI editorial
MixGuard: Towards Detecting and Understanding Mixer Laundering on Ethereum
The core problem
Mixers such as Tornado Cash and Railgun protect user privacy by concealing the link between deposit and withdrawal addresses. However, this same property is abused to launder illicit funds. Existing anti-money laundering (AML) research does not specifically target mixer laundering, while mixer research focuses on deanonymization rather than identifying laundering-related transactions. Public reports remain fragmented, leaving no public case-level dataset for systematic measurement and detection.
To fill this void, the authors present the first comprehensive study of mixer laundering on Ethereum. They construct **MixLaunder**, the first public case-level dataset of mixer laundering, covering 27 cases involving Tornado Cash and Railgun from 2020 to 2025. The dataset labels 9,300 laundering-related transactions with case identities and observable upstream and downstream fund flows, including deposits totaling approximately \$1.1 billion. By comparing these transactions with background mixer usage, the authors identify five common strategies, showing that laundering evidence spans complementary behavioral and fund-flow contexts, while same-case activity is locally tight but weakly con
Innovation
Under strict case-level holdout evaluation, MixGuard outperforms representative baselines. Key quantitative results include:
- **Detection Precision:** 97.89\%
- **Group Purity:** 98.73\%
- **Coverage:** The top ten groups cover 95.09\% of each case's transactions on average.
These results demonstrate that MixGuard effectively detects laundering-related transactions and groups them into coherent cases. The high group purity indicates that the grouping is accurate, with minimal mixing of transactions from different cases. The coverage metric shows that the top ten groups capture the vast majority of each case's transactions, making the system practical for real-world deployment.
The dataset analysis revealed five common laundering strategies, showing that laundering evidence spans complementary behavioral and fund-flow contexts. Same-case activity is locally tight but weakly connected across bursts, which poses challenges for detection. The analysis also uncovered coverage gaps in mixer-side risk screening and representative deanonymization heuristics.
Why it matters
The paper provides the first systematic measurement and detection study of mixer laundering on Ethereum. The MixLaunder dataset fills a critical gap by providing a public, case-level resource for researchers and practitioners. The identification of five common laundering strategies offers insights into how illicit actors abuse mixers.
The finding that same-case activity is locally tight but weakly connected across bursts suggests that detection systems must consider both local and global patterns. MixGuard addresses this through its tri-view representation learning, which captures complementary behavioral and fund-flow contexts.
The coverage gaps in mixer-side risk screening and deanonymization heuristics highlight the need for improved tools. MixGuard's high precision and group purity make it a promising solution for transaction-level detection and case-aware grouping.
Future work could extend the dataset to other mixers and blockchains, and explore adversarial robustness. The authors also note ethical considerations in handling sensitive data.
In summary, this work advances the understanding of mixer laundering and provides a practical detection framework, contributing to the broader fight against financial crime in decentralized ecosystems.
Who should read this
Opening member content…