The rapid adoption of 5G networks, space networks, and Internet of Things (IoT) technologies in healthcare has significantly expanded the attack surface for cybersecurity threats. This evolving landscape demands robust defense mechanisms that can anticipate and neutralize sophisticated cyber-attacks. Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) technologies are pivotal in developing such advanced cybersecurity solutions. A critical factor influencing the effectiveness of these AI/ML models is the quality and diversity of the datasets used in their training. This paper presents a systematic review of various datasets used for AI/ML-based cybersecurity model training across multiple domains, with a focus on 5G networks, IoT healthcare, and space networks. By employing a structured Goal-Question-Metric (GQM) methodology and Quasi-Gold Standard (QGS) validation, we assessed the characteristics, applications, and limitations of real, synthetic, and hybrid datasets in enhancing cybersecurity measures. The review identifies key trends, gaps, and future research directions, highlighting the need for more diverse datasets, standardized benchmarks, and privacy-preserving techniques. Our findings offer insights into improving the resilience of AI/ML models for cybersecurity, guiding the development of more effective and adaptable defense strategies across emerging network technologies.