A REVIEW OF CYBERSECURITY THREATS TO AI-BASED DISEASE DETECTION SYSTEMS IN HEALTHCARE

Artificial intelligence (AI) is increasingly used in healthcare to assist disease detection and diagnostic decision-making across medical imaging, electronic health records, and biosignal analysis. While AI offers improved diagnostic accuracy and efficiency, it also introduces unique cybersecurity vulnerabilities that can compromise patient safety and clinician trust. This review examines the landscape of cybersecurity threats targeting AI-based disease detection systems, including data poisoning, adversarial attacks, model extraction, and privacy leakage. Threats are analyzed across all stages of the AI pipeline—from data acquisition and preprocessing to model training, deployment, and clinical integration. The study further surveys defense and mitigation strategies, such as adversarial training, privacy-preserving methods, pipeline monitoring, and organizational safeguards, highlighting their effectiveness and limitations in healthcare contexts. Key challenges are identified, including the need for end-to-end security evaluation, standardized benchmarks, regulatory compliance, and integration of explainable AI to maintain clinical trust. By consolidating existing knowledge and mapping cybersecurity threats to clinical impact, this review provides a framework for researchers, clinicians, and policymakers to develop secure, reliable, and deployment-ready AI-based disease detection systems.