PTT異常帳號偵測

Abstract

隨著資訊科技日益發達與網路技術普及化下，越來越多人開始使用線上社群媒體。隨著線上社群媒體的普及化，越來越多的有心人士運用它的匿名性及社群影響力，從事網路犯罪行為，例如：網路詐欺、散布惡意程式、散布假消息、假新聞……等。這些犯罪大部分都與異常帳號密不可分，惡意使用者藉由它們隱藏自己的身分，博取其他人的信任及避免執法人員的追查。根據研究指出，Facebook約有8,800萬活耀的異常帳號，Instagram則約有 9,500萬個異常帳號。 目前國外已有許多關於惡意使用者的相關研究，然而國內較為缺乏相關的研究，並且近年國內不斷遭受假消息與假新聞的攻擊，這些都與他們操控的帳號有關。因此，本研究以國內大型論壇之一的PTT為目標，根據帳號有限的公開資料，分析PTT正常與異常帳號特徵的差異，並且建立一套分類器，提升PTT偵測異常帳號的效率。 本實驗分析PTT帳號的特徵後，發現異常帳號在使用者名稱上，並沒有明顯較高的複雜性、隨機性。然而，在活動時間的部分，異常帳號與正常帳號確實具有不同的特徵。最後，本研究將以帳號申請機制、運作模式及使用者族群的角度，分析為何PTT異常帳號會有上述特徵。

With the development of information technology and the popularization of Internet technology, more and more people are using online social media. With the popularization of online social media, more and more people use its anonymity and community influence to engage in cybercrime, such as fraud, dissemination of malicious programs, misinformation or fake news…etc. Most of these crimes are strongly related to abnormal accounts. Malicious users hide their identities by using these accounts, gaining the trust of other users and avoiding the investigation by law-enforcement officials. According to recent research, Facebook has about 88 million abnormal accounts, and Instagram has about 95 million abnormal accounts. There are a number of researches on malicious users in foreign countries, but there are few related researches in Taiwan. In recent years, Taiwan has been continuously attacked by misinformation and fake news, which are disseminated by these accounts. Therefore, our study focuses on PTT, one of the large forums in Taiwan. We analyze the differences in the characteristics between normal and abnormal accounts with limited user information, and establish a classifier to improve the efficiency of detecting abnormal accounts in PTT. By analyzing the characteristics of the PTT account, our study found that the abnormal account has no significantly higher complexity and randomness in the user name. However, the abnormal account has different characteristics from the normal accounts in the activity. Finally, our research analyzes the reasons of this outcome from the perspective of account registration system, social media features and user groups.