A practical medical image cryptosystem with parallel acceleration

Abstract

Modern medical imaging techniques, e.g., CT and MRI, can produce many privacy-sensitive images, which are usually stored or transmitted through Picture Archiving and Communication Systems (PACS). To prevent image disclosure, various encryption algorithms for protecting medical images have been introduced, such as chaotic encryption with satisfying security. Considering a practical application, we observe that medical images are usually generated in a batch manner for a particular patient. Yet, existing encryption algorithms are mostly designed for a single image, resulting in low encryption efficiency in a batch setting. Specifically, when encrypting multiple images, a qualified algorithm should be sensitive to any tiny changes of an image, i.e., spreading encryption influence of this change to the whole image set. Besides, parallel encryption naturally adapts to the batching images for speeding up, which should be considered. Thus, we aim at medical scenarios and propose a fast-and-practical cryptosystem by employing chaotic encryption and parallel computing. Our idea builds on a “permutation-substitution” architecture of chaotic encryption. In the permutation stage, the batch images are divided into multiple groups. Next, multi-threads can generate permutation coordinates parallelly in each group for saving time, in which multiple images are shuffled at one go. In the substitution stage, encryption over neighboring images via extended cipher-block-chaining (CBC) mode guarantees desirable diffusion in the whole image set. Our solution not only saves encryption time largely but also strengthens the individual’s privacy. The time complexity shows superior efficiency to counterparts, say, speeding up permutation by $\sim 60\%$. And the results of security analyses confirm that our cryptosystem can resist various attacks.