decluster

Evidence-weighted de-anonymization of Bitcoin transactions. It reads how each transaction was built (wallet fingerprints) and how its amounts partition (subtransaction structure), scores every signal as bits of evidence, and clusters coins by owner.

Two coins spent in the same transaction are normally assumed to share an owner. Because the evidence here is signed, the clustering can instead keep them apart when their fingerprints and amounts say they belong to different owners — undoing the false link a collaborative transaction deliberately plants (its whole purpose is to make an analyst merge two people into one).

• Fingerprints reveal which wallet built a transaction. Every wallet leaves quirks in how it constructs a transaction — nSequence values, script types, signature grinding, and more. Do those quirks actually identify the wallet? Taking address reuse as the same-owner label (two transactions spending the same address are the same wallet), the measured fingerprint bits rank a same-wallet pair of transactions above a random pair 93.3% of the time (AUC 0.933) on 166k real mainnet transactions. Shuffle the labels and it drops to 0.50 (a coin flip) — so the 0.933 is real signal, not an artifact.

• The shape of the payment graph reveals owners too. Independently of who-spent-with-whom, the structure of the graph (who pays whom) betrays common ownership — the same effect that de-anonymized social networks (Narayanan–Shmatikov). Across five eras (2012–2024), payment-graph structure alone predicts whether two addresses share an owner, ranking same-owner pairs correctly 0.95–0.97 of the time at one hop on the clean eras, and 0.97–1.00 across all five eras by four hops (1.0 = perfect, 0.5 = chance; the churny 2013 slice starts near chance at one hop and needs the deeper hops).

• It survives a transaction built to fool it. On a real transaction deliberately constructed to merge two owners into one (the false link from above), the method keeps them apart: the amount structure alone re-partitions them into the correct two owners, and the fingerprints independently agree — recovering the answer that a merge-only clustering gets wrong.

Layout

• decluster/ — the method: extractors, library, combiner, cluster (engine: cluster_refined), graph_deanon
• PAPER.md — the manuscript; results/ — reproducible outputs; catalog/, bigquery/

Every number is reproducible. MIT — see LICENSE.