An Intelligent Cyber Threat Prevention: A Systematic Survey of AI-Based Approaches and Open Challenges

Abstract

The increasing complexity and frequency of cyber-attacks have exposed the critical limitations of traditional signature-based security systems.Artificial Intelligence (AI) encompassing machine learning, deep learning, large language models federated learning, and reinforcement learning has emerged as a transformative paradigm for modern cyber defense. This survey critically examines significant research contributions highlighting AI applications across intrusion detection, malware analysis, phishing detection, ransomware mitigation, distributed denial-of-service prevention, and network security. A structured taxonomy of AI-driven cyber security techniques is presented, supported by analysis of widely adopted evaluation datasets including NSL-KDD, UNSW-NB15, CIC-IDS2017, and CICIoMT2024 which are assessed in terms of their scope, attack coverage and applicability to real-world deployment scenarios. Selected Pakistani criminal investigations are examined to demonstrate the operational role of AI-assisted digital forensics in judicial proceedings. The survey further identifies unresolved research challenges, including adversarial robustness, dataset generalizability, and the absence of explainability standards for court-admissible AI evidence, and outlines prospective directions to advance the development of intelligent, resilient and legally cyber security systems.