Automated Detection and Triage of Habitual Toxic Accounts Using Deep Learning

B. Jeeshitha
Dept. of CSE(AI&ML)Sri Venkateswara Collegeof Engineering Tirupati,India
bandijeeshitha555@gmail.com
K. Sudheer
Dept. of CSE(AI&ML)Sri Venkateswara Collegeof Engineering Tirupati,India
Sudheerkalugotla@gmail.com
V. Lakshmi Sameera
Dept. of CSE(AI&ML)Sri Venkateswara College ofEngineering Tirupati, India
Sameeravobulapu@gmail.com

E.Lasya
Dept. of CSE(AI&ML)Sri Venkateswara College ofEngineering Tirupati, India
etterilasya@gmail.com
Dr. D. Esther Rani
Associate professor of CSE (AI&ML)Sri Venkateswara College ofEngineeringTirupati, Indi

Abstract- This paper presents an intelligent system fordetecting and moderating toxic content in online platforms using artificial intelligence. With the rapid growth of user-generated content, manual moderation has become inefficient and unreliable. The proposed system integrates a deep learning-based toxicity detection mechanism using the Google Gemini API to analyze textual content in real time. Each post is classified into different toxicity levels such as safe, low, medium, and high. Based on these levels, a point  based mechanism is used to track user behavior overtime. Users who repeatedly generate toxic content are identified as habitual offenders and are automatically restricted or blocked once a predefined threshold is exceeded. The system consists of both user and admin modules, where users can create posts and view their activity, while administrators can monitor platform wide statistics and user behavior through a dashboard. The implementation uses Django for backend processing and modern web technologies for the frontend. The proposed approach reduces human intervention,a based on new patterns of toxic behavior Keywords- artificial intelligence, toxic content detection, deep learning, Google Gemini API, content moderation, user behavior analysis, online safety, Django framework.