Yohannes Tadesse
The rapid evolution of cyber threats has highlighted the limitations of conventional network security measures, underscoring the need for innovative, adaptive solutions. This study investigates the possibilities of artificial intelligence (AI)-driven security approaches to improve network security at every layer of the Open Systems Interconnection (OSI) model. By leveraging machine learning (ML), deep learning (DL), and natural language processing (NLP), the study introduces a cohesive framework for applying AI methods to address vulnerabilities at each OSI model layer. The study explored publicly available datasets, including CICIDS2017 and EMBER, in conjunction with real-world network data to train and evaluate AI models for various tasks, including anomaly detection, intrusion detection, malware classification, and phishing detection. The results demonstrate significant improvements over traditional security approaches, with AI-powered models achieving 90-97% accuracy in anomaly detection, 90-94% F1-score in intrusion detection with the Random Forest model, and 95-99% accuracy in malware classification. The study underscores AI's capability to analyze intricate patterns, adapt to emerging threats, and deliver immediate threat detection and response. Nonetheless, issues regarding data quality, computational complexity, and adversarial attacks have been identified as critical areas for further investigation. The results highlight the need for a comprehensive, flexible network security strategy that leverages AI to address connections across the OSI layers. This study adds to the growing body of knowledge on AI-powered cybersecurity and offers practical guidance for organizations seeking to enhance their security footprint in an increasingly connected environment.