Sarcouncil Journal of Engineering and Computer Sciences

Abstract: Cyber-attacks on hospitals continue to escalate in sophistication and frequency, threatening patient safety, clinical continuity, and data integrity. This study provides a structured review of time-series forecasting methods for hospital cybersecurity, emphasizing their value in anticipating attacks rather than responding after compromise. It examines statistical, machine learning, deep learning, and hybrid models, assessing their suitability against hospital specific challenges such as sparse and bursty incident data, high non-stationarity, and strict privacy regulations. The review analyzes internal data sources, including incident logs, IDS/IPS alerts, SIEM outputs, and network telemetry, alongside external threat intelligence used as exogenous predictors. Key limitations involve inconsistent labeling, limited dataset size, and evolving attacker behavior. The study concludes that forecasting can significantly enhance preparedness and resilience in healthcare, provided models are healthcare-tailored, context-aware, interpretable, and supported by privacy-preserving data-sharing mechanisms.