"You Know What to Do": Proactive Detection of YouTube Videos Targeted by Coordinated Hate Attacks

Video sharing platforms like YouTube are increasingly targeted by aggression\nand hate attacks. Prior work has shown how these attacks often take place as a\nresult of "raids," i.e., organized efforts by ad-hoc mobs coordinating from\nthird-party communities. Despite the increasing relevance of this phenomenon,\nhowever, online services often lack effective countermeasures to mitigate it.\nUnlike well-studied problems like spam and phishing, coordinated aggressive\nbehavior both targets and is perpetrated by humans, making defense mechanisms\nthat look for automated activity unsuitable. Therefore, the de-facto solution\nis to reactively rely on user reports and human moderation.\n In this paper, we propose an automated solution to identify YouTube videos\nthat are likely to be targeted by coordinated harassers from fringe communities\nlike 4chan. First, we characterize and model YouTube videos along several axes\n(metadata, audio transcripts, thumbnails) based on a ground truth dataset of\nvideos that were targeted by raids. Then, we use an ensemble of classifiers to\ndetermine the likelihood that a video will be raided with very good results\n(AUC up to 94%). Overall, our work provides an important first step towards\ndeploying proactive systems to detect and mitigate coordinated hate attacks on\nplatforms like YouTube.\n