Retrospective registration of a systematic review protocol: AI-Based Intrusion Detection and Prevention: A Systematic Review and Adversarial-Validity Audit of Benchmark Datasets

Abstract

Retrospective registration of a systematic review protocol. This review was conducted and submitted prior to registration; no prospective registration was performed. The review synthesizes AI-based techniques for intrusion detection and prevention, drawing on 71 peer-reviewed studies published between 2019 and 2025, selected following the PRISMA 2020 protocol. It organizes the reviewed methods into seven categories — traditional machine learning, deep learning, hybrid intelligence frameworks, blockchain-based systems, reinforcement learning, ensemble learning, and explainable AI — and compares their reported performance across more than twelve attack types using accuracy, precision, recall, F1-score, false-positive rate, and latency where available. It also contributes a provenance-and-validity audit of the benchmark datasets used across the corpus. The protocol below was reconstructed from the methods actually used in the review and is registered retrospectively for transparency and reproducibility.