Enhanced lorenz-based medical image encryption with sparse variational gaussian process modelling

The protection of sensitive medical data has become increasingly critical due to the rapid development of telemedicine and internet-based healthcare facilities. Therefore, the transfer of diagnostic images through open networks poses a significant challenge to the privacy and integrity of patients’ health records. In this paper, a medical image encryption scheme utilising the enhanced Lorenz system and the machine learning method, Sparse Variational Gaussian Process (SVGP), is proposed. The encrypted system includes steps of pre-initial scrambling, two-level diffusion, and a final permutation process. The initial scrambling relies on quantised chaotic keys produced by the enhanced Lorenz scheme and SVGP. The second round of encryption is performed using diffusion units 1 and 2, which utilise the keys of the enhanced Lorenz followed by the permutation. The enhanced Lorenz system and SVGP unit is significantly stronger in image encryption process, because of its rich hyperchaotic behaviour, high sensitivity to initial condition and large key space of 2325. Moreover, the enhanced Lorenz hyperchaotic generator is also implemented on STM32 embedded hardware to demonstrate its practical applicability. The proposed image encryption algorithm is vulnerable to a high level of security examination and yields an average entropy of 7.9993, with a near-zero correlation between neighbouring pixels. The Number of Pixel Change Rate (NPCR) and Unified Average Change in Intensity (UACI) are 99.6087 and 33.4586, respectively. The experimental evidence shows that the given approach is highly effective at ensuring confidentiality and integrity, enabling the safe sharing of medical imaging data.