結合社會網絡分析與資料探勘技術應用於網路拍賣詐欺偵測之研究

Combining Social Network Analysis and Data Mining Techniques for Internet Auction Fraud Detection

陳育騏(Yu-Chi Chen)

指導教授 ： 邱昭彰

共同指導教授 ： 李婷(Ting Lee)

元智大學/資訊學院/資訊管理學系/碩士(2007年)

摘要

網路拍賣是少數幾個能透過網際網路成功經營的商業模式。由於網路拍賣的本質，如高度匿名性、寬鬆的交易限制、很低的進入或離開成本等，使得欲詐欺者有利可圖，並在網路拍賣活動中設下騙局以進行詐欺行為。不可否認地，在買賣雙方資訊不對稱及缺乏直接讓買方驗證商品真偽的情況下，在交易完成之前，買方無法確認賣方真偽及其所賣的商品特性，這使得買方很容易成為網路拍賣詐欺的被害者。因此，為了使買賣雙方進行交易前可先行簡單過濾潛在詐欺者，以促進網路拍賣交易安全及正常化，提供網路拍賣平台業者與相關政府單位作進一步網路拍賣詐欺防治或偵查參考。本研究從拍賣網站公布之黑名單及相關訊息中蒐集帳號交易資料，透過社會網絡分析先行由所有交易帳號中過濾出異常交易帳號群組，進而再利用資料探勘技術，辨別詐欺帳號之特徵與規則。最後，交易帳號可分類為正常交易、異常交易與詐欺帳號三種不同類型並探討各類型之交易行為。本研究發現，不同交易帳號類型確實有不同的特徵與規則，其中異常交易帳號存在循環交易情況，但交易次數不高，本身在評價上無不良記錄；在詐欺帳號中，與異常交易關係相同存在循環交易，當評價在短時間內累積至一定正面評價後，負面評價隨之快速攀升，且詐欺帳號較少扮演網絡中媒介的角色。

關鍵字

網路拍賣；網路拍賣詐欺；社會網絡分析

並列摘要

Internet auction is one of few successful new business models. Owing to the nature of Internet auction, e.g. high degree of anonymity, relaxed legal constraints, and low costs for entry and exit, etc…, fraudsters are easily to setup scams in auction activities. An undeniable fact is that because of the information asymmetry between sellers and buyers and the lacking of immediately examination of the merchandise, the buyers can’t verify the seller the merchandise until after the transaction is completed. This makes buyers become victims of network auction fraud easily.
Thus, in order to promote the safety and normality of Internet auction, to assist buyers verifying sellers, so we need to provide the Internet auction platform and government to prevent or detect Internet auction fraud. This research collected account data of blacklist and relative information proclaimed by Internet auction platform. These unusual accounts were filtered by Social Network Analysis from all accounts’ data, then we using data mining techniques to discover the attributes and patterns of fraudsters. These accounts can be categorized into three parts including usual accounts, unusual accounts and accounts of fraudster, then we discuss the behaviors of transactions of these three different categories. This research found that different kinds of accounts have different characteristics and patterns indeed. Within unusual accounts, there are circle transactions, lesser times of transactions, and no negative reputations. Within fraud accounts, there are also circle transactions, positive reputation accumulated a higher scores, negative reputation which increase suddenly, and these fraud accounts less play an intermediary roles in a transaction network.

並列關鍵字

Internet auction；Internet auction fraud；Social network analysis

參考文獻

34. Wang, J.C., Chiu, C.C. and Ker, H.Y., “Detecting Online Auction Fraud of Reputation Inflation through Social Network Structures Embedded in Transaction Records,” Journal of Information Management (Taiwan), Vol. 12, No. 4, pp. 144-184, 2004.

2. Ba, S., Whinston, A.B., and Zhang, H., “Building Trust in Online Auction Markets through an Economic Incentive Mechanism,” Decision Support Systems, Vol. 35, No. 3, pp. 273-286, 2003.

3. Bierman, H.S. and Fernandez, L. Game Theory with Economic Applications, Mass: Addison-Wesley, 1998.

6. Bywell, C.E., and Oppenheim, C., “Fraud on Internet Auction,” Aslib Proceedings, Vol. 53, No. 7, pp. 265-272, 2001.

7. Chae, M., Shim, S., Cho, H., and Lee, B., “An Empirical Analysis of Fraud Detection in Online Auctions: Credit Card Phantom Transaction,” 40th Annual Hawaii International Conference on System Sciences (HICSS’07), pp. 155-164, 2007.