MULTI-MODEL FEATURES REPRESENTATION-BASED CONVOLUTIONAL NEURAL NETWORK MODEL FOR MALICIOUS WEBSITE DETECTION

Abstract

The rapid expansion of internet services and web applications has led to a significant increase in malicious websites that pose serious threats to users and organizations. These websites are commonly used for phishing, malware distribution, identity theft, financial fraud, and other cybercriminal activities. Traditional detection methods based on blacklists and handcrafted rules often fail to identify newly emerging and sophisticated malicious websites due to their dynamic nature. This paper presents a Multi-Model Features Representation-Based Convolutional Neural Network (CNN) model for malicious website detection. The proposed framework integrates multiple feature representations, including URL-based characteristics, webpage content information, domain-related attributes, and behavioral features, to provide a comprehensive analysis of website legitimacy. These heterogeneous features are transformed into unified representations and processed through a CNN architecture capable of automatically learning complex patterns associated with malicious behavior. The convolutional neural network effectively extracts discriminative features and improves classification accuracy while reducing dependency on manual feature engineering. Experimental analysis demonstrates that the proposed model achieves superior detection performance, higher accuracy, and lower false positive rates compared to conventional approaches. The framework offers an intelligent, scalable, and efficient solution for enhancing cybersecurity and protecting users from web-based threats.