DETECTION OF CHILD PREDATORS CYBER HARASSERS ON SOCIAL MEDIA

Abstract

The rapid proliferation of social media platforms has facilitated communication but simultaneously created avenues for child predators and cyber harassers to exploit vulnerable users, particularly minors. This paper proposes an intelligent detection system combining Natural Language Processing (NLP) and machine learning algorithms—including Linear SVC, Logistic Regression, Random Forest, Naive Bayes, KNN, and LSTM—to classify user messages as Normal, Harassment, or Predatory with high accuracy. The system achieves a top accuracy of 93.5% with an LSTM deep learning model trained on a TF-IDF feature space derived from labeled social media conversation datasets, and is deployed via an interactive Streamlit dashboard with real-time alert mechanisms