線上拍賣(Online Auctions)提供消費者便利的平台，使交易的進行不再受限於時間與地點，因此創造了驚人的營收。但龐大的商機也吸引了許多詐騙者混雜其中，以各種手法詐騙沒有經驗的消費者，讓此便利的平台蒙上陰影。爲了降低詐騙情事的發生，拍賣網站大多提供簡易的名聲管理系統來協助使用者挑選交易對象。由於設計過於簡單，並無法收到預期的防範效果。學者們雖有提出許多詐騙偵測方法，但其效能仍有改進空間。此外，詐騙偵測的重點不應只為了找出現行犯，應有更積極的作為，讓潛伏中的詐騙者受到監控、嚇阻甚至導正。有鑑於此，本計畫將以線上拍賣詐騙偵測、類型分析與防治策略為主軸，發展有效且可行的解決方法，預計之工作重點如下: (1) 以系統化方式建構一組有效的詐騙偵測屬性集，建立低成本、高效能的偵測模型，發展重點包含屬性的蒐集、建構與選擇，並期望能以自動方式化定期更新其內容。(2)為詐騙者進行分類，希望能設計出因應策略，在關鍵時間點以不同方式介入，以積極方式防止詐騙的發生。 Online auctions provide a convenient trading platform for consumers. Because trades through online auctions are not constrained by the time-zone and the physical location of traders, the avenue obtained from such online auctions is increasing dramatically in the recent years. However, this also attracts a lot of fraudster to join the online auctions. In view of this, the auction sites have provided simple reputation management systems to help the traders prevent form fraud. Unfortunately, such a simple design does not perform very well in fraud prevention. Thus, researchers developed a lot of fraud detection method to help the user discover the fraudsters before trading. Albeit effective, these proposed detection methods are usually relied on the experiences of human experts to construct the feature set for fraud detection. It is obviously not proper for developing an automatic detection framework. To cope with the above issues, this project tried to develop a series of methods to perform a low-cost but high-accuracy fraud detection. The main topics of the project are: (1) discovering a compact attribute set by feature selection and construction techniques: in particular, we hope the attribute set can be evolved with time automatically, (2) categorizing the fraudsters and developing the fraud-proof strategies for users: this topic is especially important for the auction sites and traders because it provides the capability to stop the fraud before it occurs.