- 學位論文
針對北高市長選舉的網軍分身偵測
Detecting Cyber Army's sock puppet for Election Campaigns
摘要
在過去傳統的選舉中，候選人主要透過廣告看板、報章雜誌及電視等來宣傳自己；近年來，由於智慧型手機的普及以及通訊技術的進步，讓人們可以輕易透過網路來分享及取得訊息，甚至自我行銷，當然對於政治人物來說也不例外。在2014年台北市長的選舉中，柯文哲成功的證明網路聲量可以轉換成選票。從此，網路就成了所有候選人的必爭之地，而在網路戰中，最廣為人知的便是利用網軍來操作議題，除了擁護支持的候選人之外，有時更是會針對競爭的候選人做出攻擊。 在本篇論文研究中，PTT是我們分析的社群媒體，其中在八卦板每天都有數萬名網友在上面討論各式議題，特別是在選舉期間，政治相關的議題更是熱目的主題。而文章的推文數越高，往往能吸引更多使用者的注意，因此為了能讓文章得到更多的關注，網軍會有所謂帶風向的行為。為了成功帶動風向，單靠一個帳號是不夠的，所以網軍會利用許多的分身帳號來推文，營造出許多人都支持或反對某一人事物的假象。 我們籍由大數據鳥瞰的方式來分析所有使用者的行為，並透過共同出現率來鎖定較常一起出現的使用者，接著再分析每群使用者間的推文時間，以及推文傾向，留下推文時間間隔較短以及推文傾向雷同的使用者群。最後再以人工的方式觀察我們所鎖定的使用者群的發文時間熱點及推文內容，以驗證我們的方式為有效找出網軍分身的方法。
並列摘要
In traditional election campaigns, candidates promote themselves via radio, television, billboard, magazines, etc. In recent years, duo to the growing popularity of mobile devices and the advances in communication technology, people can communicate, share information with others and even self-promotion via the internet even for politicians. In the mayoral election of 2014, Wen-Je Ko has proven that the volume of internet posts can turn into real votes. Since then the internet has been a battlefield for candidates. In our study, we select PTT Gossiping board as our data source. There are tens of thousands of people discussing various topics every day, and the political topics are very popular especially during the election. The more comments an article gets, the more attention it draws. In order to draw attention, cyber army’s goal is to control or change the way a matter is considered. To achieve this goal, a single account is far from enough, so cyber army will have so called “sock puppet” to leave comments for certain articles to exercise spin control. We use big data analytics on user behavior, by using co-occurrence to spot users who often participate in the discussion of the same post together. Then we filter users group with short comment time gasp and the consistency of article preference, to find cyber army’s sock puppet. To verify the outcome, we utilize heat map and word cloud.