CYBERSECURITY RISK PREDICTION USING HYBRID AI MODELS IN FINANCIAL AND HEALTHCARE SOFTWARE SYSTEMS

Abstract

Financial and healthcare software systems are among the most targeted sectors for cyberattacks due to the high value of sensitive data, strict regulatory requirements, and the increasing adoption of cloud-native and interconnected digital infrastructures. Traditional security tools struggle to anticipate risks that evolve rapidly, such as zero-day exploits, ransomware, and insider threats, especially in environments where large volumes of structured and unstructured data must be secured in real time. To address these challenges, this study explores the use of hybrid artificial intelligence (AI) models—combining machine learning classifiers with generative and deep learning approaches—for predicting cybersecurity risks in critical financial and healthcare applications. Hybrid AI systems integrate the strengths of discriminative models for anomaly detection with the adaptive capabilities of generative models for simulating novel attack scenarios, thereby improving accuracy, reducing false positives, and enhancing resilience against previously unseen threats. Empirical evaluations conducted on publicly available security datasets, including UNSW-NB15 and healthcare intrusion logs, demonstrate that hybrid AI models can achieve over 95% detection accuracy while maintaining lower computational overhead compared to standalone deep learning architectures. Moreover, the proposed approach aligns with compliance frameworks such as HIPAA and PCI-DSS by incorporating explainability and traceability into the prediction pipeline. These findings highlight the transformative potential of hybrid AI for risk prediction in high-stakes domains, offering a pathway toward more reliable, proactive, and regulation-compliant cybersecurity defenses in financial and healthcare systems.