Capturing Cross-Platform Interaction for Identifying Coordinated Accounts of Misinformation Campaigns

Abstract

Recent years have witnessed the increasing abuse of coordinated accounts on multiple social media platforms. Such accounts are usually operated by misinformation campaigns to manipulate the public opinions on different platforms jointly. However, existing methods mainly focus on detecting such accounts by capturing the coordinated activities within a single platform. As a result, their performances are limited as they can not make use of the information from other platforms. In this work, we propose that capturing cross-platform coordinated activities can bring a significant boost to identifying the accounts operated by misinfromation campaigns. To leverage such information in a practical way, we design a novel Conditional Gaussian-distribution Basis to extract cross-platform correlation from Coordinated Activity Set, which can be easily acquired. Experimental results indicate that our methodology outperform baselines and its own variants that can not leverage cross-platform information.

Yizhou Zhang

Ph.D Candidate in Computer Science

My research interests include machine learning and its application on social media.