由於網拍的可見商機與低進入成本，使得國內外網拍平台皆成為網路詐欺犯罪的溫床，其中常見的手法是大量建立帳號利用這些帳號建立假交易以增加彼此評價分數，雖不是直接造成交易方財產損失的詐欺方式，卻是實行詐欺前重要的取信手段。本研究針對此種詐欺共犯群體，提出一個模組化的詐欺共犯群體偵測流程，以彌補網路拍賣評價系統之不足。將流程分為前處理、分群、合併、過濾、比對五階段，其中在分群階段以社會網路分析分群方法之K-Plex與N-Club為主要方法。而系統各階段的設計過程中將以詐欺共犯群體之組織特性為基礎設計與調整各模組的實作方式。本研究從露天拍賣網站收集真實線上拍賣交易評價資料，並使用所提出之階段性的模組化偵測流程進行實做與驗證。 Due to the great profit and low entrance cost of online auction, there are a lot of online fraud cases in online auction sites. One of the most common fraud method is to establish several auction account, and using these account to have transactions with each other then increase their reputation. The research focus on this kind of the collusive fraud group and propose a process to detect collusive group in order to complete the flaws of reputation system in the online auction. We divide the process into five steps: preprocessing, classification, merge, filter, and validation. In the step of classification, we use the K-Plex and N-Club approach of social network analysis as the primary method. Otherwise, in the design process of each step, we design and adjust the implementation of each module based on organizational characteristics of the collusive fraud group. In this research, we collect the real reputation data of online auction from "www.ruten.com.tw", and use the detective process we propose to do implementation and validation.
