Volume 1 - Issue 4, November - December 2025

Cybersecurity plays a critical role in safeguarding digital infrastructure from increasingly advanced and sophisticated cyber threats. Traditional intrusion detection systems typically rely on predefined rules, known signatures, and static behavioural patterns. Although effective against known attacks, these systems often fail to recognize novel, zero-day, polymorphic, and rapidly evolving attack vectors. As cyber adversaries adopt more complex and intelligent techniques, there is a growing need for adaptive and autonomous security solutions. To overcome these limitations, Artificial Intelligence (AI)–based anomaly detection has emerged as a powerful and transformative approach in modern cybersecurity. AI-driven anomaly detection systems continuously monitor network behaviour, learn normal By leveraging machine learning, deep learning, and statistical modelling, these systems can detect subtle, previously unseen abnormal patterns that traditional systems often miss. Furthermore, AI enables real-time threat analysis, automatic classification, and early alerting, helping organizations respond to intrusions be This paper provides a comprehensive study of AI-based anomaly detection in cybersecurity, covering key algorithms such as supervised, unsupervised, and reinforcement learning models; widely used benchmark datasets It also discusses key performance evaluation metrics including accuracy, precision, recall, F1-score, and detection latency. In addition, real-world applications such as intrusion detection, malware the practical relevance of AIbased systems. As cyber threats continue to evolve, the integration of AI with modern security technologies are expected to further improve detection precision and efficiency. AI systems will increasingly support automated response mechanisms, reducing the burden on human analysts.