Artificial Intelligence–Based Zero-Day Attack Detection and Proactive Mitigation Strategies: A Literature Review

Abstract

Zero-day attacks pose a significant challenge in cybersecurity, as they exploit previously unknown vulnerabilities and circumvent traditional signature-based defenses. This work proposes an artificial intelligence–driven framework to identify, anticipate, and mitigate emerging zero day threats. By combining machine learning, deep learning, and behavioral analytics, the framework detects abnormal system behavior and identifies malicious activity in real-time. Unsupervised anomaly detection, sequence‑based neural models, and AI‑supported threat intelligence are employed to forecast potential weaknesses and predict exploit trajectories before active exploitation occurs. The framework further enables automated countermeasures such as virtual patching, adaptive response strategies, and dynamic risk evaluation. Observations from experimental analysis demonstrate that AI‑enabled defenses reduce false alarms, accelerate response time, and improve proactive security. Overall, the study confirms that intelligent, data‑driven systems significantly enhance resilience against previously unknown and rapidly evolving cyber threats.