Address Clustering Heuristics for Distributed Ledgers

Abstract

The paper surveys heuristic methods of clusterization in address space of public distributed ledgers. The techniques mentioned rely on collecting behavioral patterns of typical actors and some common sense. Formally heuristics are degenerate clusterization rule-based algorithms, which do not tune their parameters via learning on curated datasets. They can be treated also as persistent motifs in transaction networks. Despite its seeming sim-plicity and inability to assess correctness of the results such approach demonstrates a reasonable effectiveness and often is used as a preliminary step before applying much more sophisticated tools based on machine learning algorithms and AI. Heuristics for Bitcoin, Ethereum, Ripple, Monero and Zcash are discussed. Heuristic clusteri-zation in cross-chain setting is briefly mentioned. Cases when heuristic approach leads to incorrect results are discussed.