More than 40 openly documented heuristics and AI methods that Coinator uses for Bitcoin forensics.
Unlike many other transaction-tracing tools, Coinator deliberately relies on transparency instead of secrecy. We make no mystery of our algorithms, heuristics, and analysis procedures — on the contrary, every method used is openly documented and always reviewable.
More than 40 clearly traceable procedures — including multi-input heuristics, change-address analyses, CoinJoin and PayJoin detectors, and numerous filters — ensure maximum traceability, high trustworthiness, and court-admissible results at European data-protection standards.
We organise our classical procedures into ten categories. Each contains several specific heuristics that, combined, feed into cluster and transaction analysis — all openly documented and traceable.
Addresses that appear together as inputs in a transaction very likely belong to the same entity. The base heuristic of blockchain forensics.
Detecting the return-change output in Bitcoin transactions. Change addresses belong to the sender and extend the cluster.
Detecting collaborative transactions to avoid clustering CoinJoin participants incorrectly. Used in combination with multi-input heuristics.
Identifying classical mixing services by their typical input and output patterns.
Cross-chain swaps allow direct exchange of coins between different blockchains — without a centralised exchange and therefore without KYC. Originally built for arbitrage, they are increasingly used to obscure the origin of funds. Our Coinator recognises the typical on-chain patterns of several established protocols.
Recurring temporal patterns reveal wallet characteristics — private vs. service, region, usage type.
Analysing graph structure: who flows where, via how many intermediate hops?
Identifying the entity types behind addresses: exchanges, mining pools, payment gateways, custodians.
Tracking "contaminated" coins through the blockchain — a core concept for source-of-funds checks and AML.
Additional heuristics that address individual aspects.
Classical heuristics are augmented by modern machine-learning methods — for higher precision and fewer misclassifications.
Coinator combines classical heuristic procedures with machine-learning (ML) and data-mining (DM) methods to analyse Bitcoin addresses and transactions more precisely. The goal is to detect relationships that go significantly beyond conventional rules.
We employ neural networks (Graph Neural Networks in particular) to uncover address relationships that escape classic heuristics like multi-input or change-address analysis. The GNNs are trained on carefully curated datasets to deliver reliable network analyses in real-time production use.
Sometimes the "key" to insight lies not in the network topology itself but hidden in user behaviour. Coinator leverages this by grouping transactions with AI based on temporal relationships, sequences, changes in amounts, and recurring (sometimes atypical) usage patterns.
"Divide and conquer." This innovative approach runs ML analyses on (sub-)transaction-graphs to identify hidden (sub-)relationship networks and to derive and name entities automatically.
Finally, we combine proven classical heuristics with the AI and ML algorithms above to improve entity recognition and reduce false positives.
We're happy to explain which procedures best fit which case.
Get in touch →