온라인 공간에서 관심집단 대상 비정상 정보의 특징 분석과 탐지 = Characterization and Detection of Opinion Manipulation on Common Interest Groups in Online Communities

국문 초록 (Abstract)

인터넷 포털과 사회관계망 서비스(SNS) 등의 온라인 공간에서 사용자 간의 의견 공유가 활발해짐에 따라 이를 악용하여 특정 개인이나 집단의 이익을 위해 유포되는 비정상 정보도 증가하고 있다. 특히 비정상 정보가 정치적인 목적으로 유포되면 선거 결과뿐 아니라 다양한 사회 정책과 시민 생활에도 영향을 미친다. 이러한 비정상 정보는 불특정 다수에 대한 유포에서 시작하였으며 이들의 특성을 분석하고 탐지하기 위한 기존 연구도 이러한 불특정 다수 대상 유포에 초점을 맞추었다. 하지만 최근에는 더욱 효과적으로 영향을 미치기 위해 공통 관심사를 가진 집단(예: 부동산에 관심 있는 사람들의 모임)을 대상으로 내용과 형식을 조정한 맞춤형 정보를 유포하고 있다. 본 논문에서는 이러한 관심 집단을 대상으로 한 비정상 정보의 특성을 분석하고 이를 탐지하는 방법을 제시한다. 이를 위해 선거 전후에 10개의 공통 관심 집단에 게시된 의견을 수집하여 분석하였다. 그 결과, 각 집단에 맞춤화된 정보가 실제 유포되고 있으며 선거일이 가까워짐에 따라 점차 증가함을 보였다. 또한, 비정상 정보를 탐지하기 위한 시스템을 제안하였는데, 이 시스템은 개별 의견에서 보이는 특징뿐 아니라 의견 게시자의 전반적인 행위 및 게시자와 협력한 사용자의 특성을 종합적으로 분석한다. 제안한 시스템을 수집한 데이터에 적용한 결과 90% 이상의 정확도로 비정상 의견을 탐지하였으며 다수의 사용자가 조직적으로 비정상 의견을 유포한 정황을 발견하였다. 제안한 시스템으로 관심 집단에 게시된 의견을 주기적으로 검사한다면 비정상 정보의 유포를 더 빠르게 차단하고 영향을 줄일 수 있을 것이다. 또한, 탐지에 활용한 특징은 정치적인 목적 이외의 비정상 정보 판별에도 활용될 수 있을 것이다.

다국어 초록 (Multilingual Abstract)

As more people share their opinions in online communities, such as Internet portals and social networking services, more opinions are manipulated for the benefit of particular individuals and groups. In particular, when manipulations occur for political purposes, they influence election results as well as government policies and the quality of life. This type of manipulation has targeted the general public, and their analysis and detection has also focused on such manipulation. However, to more efficiently spread propaganda, recent manipulations have targeted common interest groups(e.g., a group of those interested in real estate) and propagated information whose content and style are customized to those groups. This work characterizes such manipulations on common interest groups and proposes method to detect manipulations. To this end, we collected and analyzed opinions posted on 10 common interest groups before and after an election. As a result, we found that manipulations on common interest groups indeed occurred and were gradually increasing toward the election date. We also proposed a detection system that examines individual opinions, their authors, and their collaborators. Using the collected opinions, we demonstrated that the proposed system can accurately classify more than 90% of manipulated opinions and that many of these opinions were posted by multiple collaborators. We believe that regular audits of opinions using the proposed system can quickly isolate manipulations and decrease their impact. Moreover, the proposed features can be used to identify manipulations in domains other than politics.

참고문헌 (Reference)

• 1 "“News articles in Daum news"
• 2 "Weka: the workbench for machine learning"
• 3 B. Liu, "Web data mining: exploring hyperlinks, contents, and usage data" Springer 2011
• 4 Y. Kim, "We can do everything with Naver IDs illegally sold at only 800 won"
• 5 B. Collins, "Twitter pulls down bot network that pushed pro-Saudi talking points about disappeared journalist"
• 6 M. Fraser, "Throwing sheep in the boardroom: how online social networking will transform your life, work and world" Wiley 2008
• 7 R. K. Garrett, "The promise and peril of real-time corrections to political misperceptions" 1047-1058, 2013
• 8 V. N. Vapnik, "The nature of statistical learning theory" Springer 2000
• 9 H. Grassegger, "The data that turned the world upside down" VICE Media Group
• 10 D. Centola, "The Spread of Behavior in an Online Social Network Experiment" American Association for the Advancement of Science (AAAS) 329 (329): 1194-1197, 2010
• 11 H. K. Lee, "South Korea takes new measures to have enough face masks domestically amid coronavirus"
• 12 S. Choe, "Shadowy church is at center of coronavirus outbreak in South Korea"
• 13 L. Breiman, "Random forests" 45 (45): 5-32, 2001
• 14 S. Choe, "Prosecutors detail attempt to sway South Korean election"
• 15 "News articles in Naver news"
• 16 "Naver cafe"
• 17 J. Lee, "Manipulation of recommendation counts by military and government agencies"
• 18 E. L. Park, "KoNLPy: Korean natural language processing in Python" 133-136, 2014
• 19 S. Shane, "Inside a 3-year Russian campaign to influence U.S. voters"
• 20 Alex Krizhevsky, "ImageNet classification with deep convolutional neural networks" Association for Computing Machinery (ACM) 60 (60): 84-90, 2017
• 21 J. Robertson, "How to hack an election"
• 22 M. B. Khalifa, "Evidental group spammers detection" 341-353, 2020
• 23 Sihyung Lee, "Detection of Political Manipulation in Online Communities through Measures of Effort and Collaboration" Association for Computing Machinery (ACM) 9 (9): 1-24, 2015
• 24 J. Ratkiewicz, "Detecting and tracking political abuse in social media" 2011
• 25 Viresh Gupta, "Detecting and Characterizing Extremist Reviewer Groups in Online Product Reviews" Institute of Electrical and Electronics Engineers (IEEE) 7 (7): 741-750, 2020
• 26 "Daum cafe"
• 27 K. Kohli, "Congress vs. BJP: the curious case of trolls and politics"
• 28 Wu Youyou, "Computer-based personality judgments are more accurate than those made by humans" Proceedings of the National Academy of Sciences 112 (112): 1036-1040, 2015
• 29 Delia Mocanu, "Collective attention in the age of (mis)information" Elsevier BV 51 : 1198-1204, 2015
• 30 Prithviraj Sen, "Collective Classification in Network Data" Association for the Advancement of Artificial Intelligence (AAAI) 29 (29): 93-, 2008
• 31 "Bulletin board posting guidelines"
• 32 Georges Dupret, "Bootstrap re-sampling for unbalanced data in supervised learning" Elsevier BV 134 (134): 141-156, 2001
• 33 S. Choe, "Ally of south Korean leader conspired to rig online opinion, inquiry finds"
• 34 G. H. Nguyen, "A supervised learning approach for imbalanced data sets" 1 (1): 1-4, 2008
• 35 R. E. Schapire, "A brief introduction to boosting" 2 : 1401-1406, 1999
• 36 R. Bond, "A 61-million-person experiment in social influence and political mobilization" 489 (489): 295-298, 2012