線上拍賣詐騙行為層出不窮，拍賣網站常用的二元名聲管理系統效能不彰，讓許多消費者平白蒙受損失。為防範此類惡行日益擴大，本研究分析線上拍賣詐騙者之交易資料，建立詐騙者偵測模型，以協助消費者儘早發覺異常交易。為達成此目標，我們首先對常見的詐騙行為進行觀察，並歸納出七項偵測指標。而後，擷取Yahoo拍賣網站被停權的詐騙者公開資料，並分別進行迴歸分析與資料分類，以建立詐騙者偵測模型。為早期發現詐騙行為，上述分析將分別根據詐騙者各階段的交易歷史來進行。實驗結果顯示，本研究提出方法在早期詐騙偵測成功率幾乎可達100%。此外，我們也建立一套詐騙查詢系統，以協助Yahoo拍賣使用者發覺潛在的詐騙者。 There are more and more deception on electronic auction. The binary reputation system for the auction website doesn’t work very well and let many consumers lose a lot. To guard against this kind of dirty trick to expand, our research analyses the transaction data of cheater on online auction, and build up a deception detection model to help consumers to find some unusual transactions as early as possible. In order to come to this goal, we are observing some common deception and conclude seven rules. Then, gather public data of cheater on Yahoo auction, and do regression analysis and data classification to build up a deception detection model. To find fraud early, these analyses are according to the step of transaction history. The result shows that an early deception detection is almost 100%. And we also build a deception querying system to help consumers finding some potential cheater.
