AutoFR: Automated Filter Rule Generation for Adblocking

• DOI: 10.48550/arxiv.2202.12872
• Publication Date: 2022-01-01
• Authors: Hieu Lê; Salma Elmalaki; Athina Markopoulou; Zubair Shafiq
• Keywords: Computer Science - Networking and Internet Architecture; Networking and Internet Architecture (cs.NI); FOS: Computer and information sciences; Computer Science - Machine Learning; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology; Machine Learning (cs.LG)

Adblocking relies on filter lists, which are manually curated and maintained by a community of list authors. Filter curation is laborious process that does not scale well to large number sites or over time. In this paper, we introduce AutoFR, reinforcement learning framework fully automate the rule creation evaluation for interest. We design an algorithm based multi-arm bandits generate rules block ads while controlling trade-off between blocking avoiding visual breakage. test AutoFR thousands show it efficient: takes only few minutes site effective: generates can 86% ads, as compared 87% EasyList, achieving comparable Furthermore, generalize new sites. envision assist adblocking in generation at scale.