Paper Title
A MULTIMODAL WEB EXTENSION FOR REAL TIME OBSCENITY DETECTION AND CONTENT MODERATION

Abstract
The rapid expansion of internet access and digital media consumption has significantly increased user exposure to obscene, explicit, and harmful online content. Such exposure poses serious psychological, ethical, and productivity-related concerns, particularly for children, students, and corporate environments. Conventional content filtering techniques relying on keyword matching or static blacklists are inadequate in handling dynamic, multimodal web content. To address these challenges, we propose an AI-driven Obscenity Blocker Web Extension designed to provide real-time, intelligent moderation of online content. The system employs a multimodal machine learning framework that integrates natural language processing for contextual profanity detection, convolutional neural networks for image and video classification, optical character recognition for extracting embedded text, and speech-to-text analysis for detecting profane audio content. The extension actively monitors webpage content using DOM observers and applies automated censorship mechanisms such as text masking, media blurring, and website blocking without disrupting user experience. In addition, the system incorporates perceptual image hashing to identify repeated circulation of obscene media and a centralized reporting mechanism that forwards verified content to a nodal agency for regulatory analysis. Guardian and employer notification features further enhance accountability by triggering alerts upon policy violations or tampering attempts. Periodic analysis of reported data enables continuous model refinement and trend monitoring. By combining real-time detection, multimodal analysis, and centralized oversight, the proposed solution contributes to creating safer digital environments while maintaining scalability and performance across diverse browsing scenarios. Keywords - Obscenity Detection, Content Moderation, Multimodal Learning, Web Extension, Computer Vision, Natural Language Processing, Online Safety