A Real-Time AI-Powered Framework for Integrated Malware, Phishing, and Security Breach Detection

Abstract

The rapidly evolving cyber threat landscape, characterized by sophisticated malware, targeted phishing campaigns, and stealthy security breaches, poses significant challenges to traditional signature-based detection systems. This paper proposes a unified, real-time framework that leverages a ensemble of machine learning (ML) and deep learning (DL) techniques to provide comprehensive threat detection. Our hybrid approach employs Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks for static and dynamic malware analysis, Natural Language Processing (NLP) with Transformer-based models for phishing email and URL classification, and unsupervised anomaly detection algorithms like Isolation Forest for identifying novel network and host-based breaches. The framework is designed to process heterogeneous data sources—including executable files, network traffic, and system logs—in a scalable pipeline. Evaluated on public datasets such as CIC-MalMem-2022, CICIDS-2017, and a phishing corpus, the proposed model demonstrates high efficacy, achieving an average F1-score of 98.2% for malware classification, 97.5% for phishing detection, and 96.8% in anomaly-based breach detection with a low false positive rate. The results underscore the potential of a consolidated AI-driven framework to enhance situational awareness and provide proactive, real-time security in modern digital environments.