AI-DRIVEN CYBERSECURITY: MACHINE LEARNING APPROACHES FOR DETECTING ZERO-DAY ATTACKS

Zero-day attacks have become one of the most persistent and damaging forms of cyberthreats because they exploit vulnerabilities unknown to security vendors and defenders. Traditional detection mechanisms that depend on signatures or predefined rules cannot identify these emerging threats in time. Artificial intelligence (AI), particularly machine learning (ML), presents a highly promising alternative by enabling security systems to recognize unusual patterns that may indicate a previously unseen attack. This research paper explores how ML-based techniques can enhance zero-day detection, reviews current research in the field, and evaluates different model types based on performance, adaptability, and practicality. Two comparative tables are included to summarize detection effectiveness and model capabilities. The study concludes that AI-driven detection frameworks outperform traditional tools by learning dynamic behavior, although challenges such as adversarial manipulation, computational cost, and lack of high-quality datasets still require careful consideration.