Bitcoin is the most popular cryptocurrency that features a distributed and decentralized mechanism. Unfortunately, some criminal behaviors which took advantage of this platform were not identified. This has discouraged many governments to support cryptocurrency. Our goal is to detect the usage of a Bitcoin address. We propose the transaction history summarization including conventional features and temporal information. We evaluated them with several supervised machine learning algorithms and found that the gradient boosting decision tree method, LightGBM, works the best.

Collaborative note: https://hackmd.io/@coscup/rJcxx7e4r