AI-Driven Techniques for Insider Threat Detection: A Comprehensive Systematic Review and Future Directions

Insider threats—malicious or negligent actors with authorized access—remain one of the hardest cybersecurityproblems to detect. This paper presents a systematic, reproducible review of AI-driven insider-threat detectiontechniques, emphasizing methodology (PRISMA-style selection), datasets and evaluation practices, model taxonomies,deployment constraints, privacy and legal considerations, and open research directions. We synthesize evidenceacross supervised, unsupervised, semi-supervised, hybrid, deep learning, and graph-based approaches and provide acurated bibliography of 100 peer-reviewed studies (2010–2025) to support reproducible follow-up studies