RELIABLE AI FRAMEWORK FOR REAL-TIME DETECTION OF PHISHING AND MALWARE IN SMS AND EMAIL

Abstract

The rapid growth of digital communication platforms such as Short Message Service (SMS) and email has brought new cybersecurity challenges. These platforms are now commonly used by cybercriminals to carry out phishing attacks, spread malware, and manipulate users through social engineering techniques. As these threats continue to evolve, traditional security methods like rule-based filters and signature-based systems are becoming less effective in identifying advanced attacks.

This study presents a reliable artificial intelligence-based framework designed to detect phishing and malware in real time across SMS and email networks. The proposed system combines transformer-based language models with graph neural networks to examine both the content of messages and the communication patterns between users. This helps in identifying not only individual threats but also coordinated cyber-attacks.

To improve trust and usability, the framework also includes explainable AI techniques, which make the system’s decisions easier to understand. It further incorporates adversarial robustness to handle manipulated inputs and federated learning to ensure user data privacy.

Experimental results using phishing and SMS spam datasets show that the proposed approach achieves higher detection accuracy and reduces false alarms compared to traditional machine learning methods. Overall, the study highlights that combining text analysis with network-level insights is an effective way to detect cyber threats early and improve the security of digital communication systems.