為提升線上拍賣交易的安全性，本研究提出一套二階段不誠實交易者的偵測機制: 第一階段力專注於詐騙偵測，第二階段則致力找出拍賣社群中的不誠實交易者(或共犯)。在詐騙偵測部分，本研究考量以『盜取帳號』為基礎的新型態詐騙，並配合詐騙者行為時序，提出六種新的偵測指標。在搜尋可疑交易者方面，本研究先以clique為基礎，在交易網路中找尋緊密交易關係，再利用這些結構建立偵測指標，標示出其中的可疑交易者。為驗證系統有效性，我們由Yahoo-Taiwan擷取實際交易資料進行實驗。在詐騙偵測部分，本研究提出方法之正確度(Accuracy)超過84%，優於Chau等人提出之十七項偵測指標[9]。對於不誠實交易者的偵測，本研究提出之指標對4 clique成員有80%以上之偵測正確度，對於3 clique成員也有75%之正確度。總結上述實驗結果，我們相信本論文之研究成果對線上拍賣詐騙防治能有實際貢獻，並有助於線上拍賣應用領域的進一步拓展。 To increase security level in online auctions, we proposed a two-stage detection mechanism for dishonest traders in online auctions. The first stage of the proposed mechanism focuses on fraud detection, and the second stage is to discover dishonesty traders or criminal colluders. However, identity theft is a new type of fraud is to cheat trading partners for grabbing money. Therefore, we proposed six indices that based the concept of clique on social network analysis to measure behavioral sequences for detecting fraudsters in stage one to discover suspects and among which the relationships are. To validate the effectiveness of our system, we also collected real transaction data from Yahoo in Taiwan to detect dishonest traders using the proposed indices. The accuracy of stage 1 experiment is 84% accuracy that is better than Chau’s indices. And the accuracy of detecting 4 cliques is over 80%, and over 75% in detecting 3 cliques. In conclusion, the results of this study could be helpful in online auction fraud prevention and implementing further application in online auctions.
