Title: How to Peel a Million: Validating and Expanding Bitcoin Clusters

Abstract: One of the defining features of Bitcoin and the thousands of cryptocurrencies that have been derived from it is a globally visible transaction ledger. While Bitcoin uses pseudonyms as a way to hide the identity of its participants, a long line of research has demonstrated that Bitcoin is not anonymous. This has been perhaps best exemplified by the development of clustering heuristics, which have in turn given rise to the ability to track the flow of bitcoins as they are sent from one entity to another.
In this paper, we design a new heuristic that is designed to track a certain type of flow, called a peel chain, that represents many transactions performed by the same entity; in doing this, we implicitly cluster these transactions and their associated pseudonyms together. We then use this heuristic to both validate and expand the results of existing clustering heuristics. We also develop a machine learning-based validation method and, using a ground-truth dataset, evaluate all our approaches and compare them with the state of the art. Ultimately, our goal is to not only enable more powerful tracking techniques but also call attention to the limits of anonymity in these systems.