ESecMedIR: Efficient and Secure Dual-level Transformer based Medical Image Retrieval Framework in the Cloud

Critical information within medical images helps in accurate diagnosis and research. Nevertheless, this increases the need for efficient storage and retrieval of information embedded in medical images. Moreover, it poses crucial challenges, such as privacy management and secure retrieval of these images. To address these challenges, this paper proposes an Efficient and Secure Medical Image Storage and Retrieval framework (ESecMedIR) leveraging cloud computing. To address privacy concerns and the challenge of balancing security with retrieval efficiency, ESecMedIR utilizes Dual-level Vision Transformer-based Hashing and Privacy Region Encryption. Sensitive regions in medical images are identified and encrypted using hyperchaos-based encryption, while hashcodes are generated and used in similar image searches for efficient retrieval. The proposed framework is tested on three standard medical datasets of Brain MRI, Chest X-ray, and Kidney CT, demonstrating a 10%–30% improvement in the retrieval accuracy over existing methods, ensuring secure and efficient image storage and retrieval management.