AI-Driven Adaptive Lightweight Cryptography for IoT Healthcare Systems: A Contextual Security Framework for Resource-Constrained Environments

Abstract

The rapid proliferation of Internet of Things (IoT) devices in healthcare systems has introduced critical security challenges, particularly in resource-constrained environments typical of African healthcare infrastructures. Traditional cryptographic solutions impose prohibitive computational overhead on low-power medical IoT devices. This paper proposes an AI-driven adaptive cryptography framework capable of dynamically selecting between the lightweight encryption algorithms RECTANGLE and ASCON based on data sensitivity, network conditions, and real-time geolocation. We trained a Random Forest classifier on the MedMC-QA medical dataset to automatically classify healthcare data sensitivity into three categories (HIGH, MEDIUM, LOW), achieving an accuracy of 92.4% ± 1.2% (95% CI) after 5-fold cross-validation. Experimental evaluation across three deployment scenarios (hospital, home, and rural area) in Douala, Cameroon, demonstrated average encryption times of 52.33 ms while ensuring robust security for sensitive medical data. Statistical analysis revealed significant performance variations across contexts (ANOVA, p = 0.007), highlighting the need for contextual adaptation. The proposed framework constitutes a practical and scalable solution for securing IoT healthcare systems in resource-constrained environments, while ensuring compliance with medical data protection requirements.