Detection of Malicious URLs Using a Hybrid Data-driven Adaptable Framework

Cybersecurity threats have increased dramatically with the growth of online activity, as malicious URLs cause major financial, privacy, and data losses. This work aims to build an intelligent system that analyzes and classifies URLs using a large dataset to detect and prevent malicious activity. The system employs advanced Machine Learning and Deep Learning models implemented with Scikit-learn and TensorFlow for accurate, real-time detection. Models including Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (BiLSTM) are trained and evaluated. The system extracts lexical, structural, and statistical URL features to enhance accuracy and adaptability. This work focuses on strengthening the concepts of cybersecurity by implementing URL-constructed threat detection management system. The model was evaluated based on several performance measures across the other state-of-the-art approaches. It was evident that CNN–BiLSTM accomplished the best results with testing accuracy of 98.6% and training accuracy of around 98.3%. This established a proof of concept towards a reliable and intelligent cyber defense system for the future.