Automatic Bitcoin Address Clustering

A white paper was released from Bitfury titled Automatic Bitcoin Address Clustering which proposes an algorithm to de-anonymize Bitcoin transactions using a combination of on-chain and off-chain information.

The first heuristic used is considered Common Spending CS. If two or more addresses are inputs of the same transaction with one output, then all these addresses are controlled by the same user. The second heuristic used is One-time change OTC based on the standard Bitcoin mechanism where the change from a transaction is returned to a new address. The paper then proposes to use tag collection from either passive web crawling or active manual analysis of Bitcoin companies and data actualization procedures. The active analysis suggest that the most common Bitcoin business companies are exchanges, marketplaces, mining pools and mixers. Some of which mostly use addresses with specific prefixes. The next stage in analysis is the use of Negative pairs. The original tag collection is considered dirty tags since they are not standardized. The negative pairs are are considered clean since they are prepared and may correspond to only one tag types. Tag types they proposed of Bitcoin organizations are mining pools, exchanges, Darknet markets, mixers, gambling, and other services.

Using the combination of on-chain and off-chain information fed into a machine learning clustering algorithm the paper is able to produce more accurate clustering of addresses than previous attempts. The use of the negative pairs helps to refine and produce more homogeneous clusters which is an improvement over purely on-chain information.

Support us and the authors of this article by donating to the following address:


Comments powered by Talkyard.