A New Bitcoin Address Association Method Using a Two-Level Learner Model

Users in the Bitcoin system adopt a pseudonym-Bitcoin address as the transaction account, making Bitcoin address correlation analysis a challenging task. Under this circumstance, this paper provides a new Bitcoin address association scheme which makes address tracing possible in Bitcoin systems. The proposed scheme can be used to warn relevant institutions to study more secure encryption algorithms to protect users’ privacy. Specifically, the important features of a Bitcoin address are extracted. After that, to reduce the computational complexity, we transform the clustering problem of addresses into a binary classification problem in which we integrate the features of two Bitcoin addresses. A novel two-level learner model is then built to analyze if the two Bitcoin addresses are belonging to the same user. Finally we cluster the addresses belonging to the same user accordingly. Extensive experimental results show that the proposed method outperforms the other address association schemes, which use deep learning models or heuristics, and can achieve an increase by 6%–20% in precision and by 10% improvement in recall.