Bitcoinは, 信頼できる中央機関の存在を必要としない金融取引システムであり, 少額の取引手数料での送金, 透明性の高い寄付および出資などの手段として注目を浴びている。
しかしながら, Bitcoinは匿名性があり, 投資詐欺などに悪用されることが問題となっている。
そこで本研究では, 投資詐欺に関連したBitcoinアドレスをWebから収集する手法を提案し, Bitcoinの取引履歴を機械学習を用いて解析する手法を提案し, さらに, 犯罪に用いられる取引をリアルタイムに検知するシステムを構築した。結果, 918個のBitcoinアドレスを発見し, 偽陽性率を3.8%に抑え, 88%の精度で検知可能なことを示した。
Bitcoin is one of the most successful decentralized cryptocurrencies to date. However, it has been reported that it can be used for investment scams, which are referred to as HYIP (High Yield Investment Programs). So far, no schemes has been proposed to detect HYIP operators' Bitcoin addresses, although it is useful from the security forensics aspect. We have proposed a novel scheme to identify HYIP operators' Bitcoin addresses by analyzing transactions history. We collected 918 HYIP operators' Bitcoin addresses from the Internet and analyzed the characteristics of transactions where the collected Bitcoin addresses are involved. Based on this analysis, we proposed a machine learning technique to classify given Bitcoin addresses into HYIP operators ones or not. By evaluating the classification performance , our best scheme achieves that 88% of HYIP addresses are correctly classified, while maintaining false positive rate less than 3.8%. We also built a web application for practice.
|