Research in the International Journal of Information and Communication Technology describes a new approach to the encryption of digital images. The method could be used to protective sensitive information, such as medical and scientific images, online. By using chaotic systems to do the work, the approach, developed by Zhengbao Cai of the College of Information Technology in Lu’an, China, improves on existing approaches.
Digital image transmission has made encryption essential for safeguarding personal data, medical records, business, political, and military intelligence. However, traditional encryption methods, such as the Advanced Encryption Standard (AES), have limitations when it comes to handling complex and dense data of the kind found in a digital image file. To work around the various problems, Cai turned to a chaotic encryption system. Such an approach uses the irregular and nonlinear dynamics of chaos theory to obscure data. The new work introduces a six-dimensional cellular neural network (CNN) that can encrypt colour more efficiently and with lower demands on computing resources than earlier chaos-based methods.
Conventional two- or three-dimensional CNNs models generate sequences of chaotic numbers that are highly unpredictable. By taking that approach to a higher dimension, Cai improves on the degree of unpredictability as well as making the encrypted output more stable when encrypting large, high-dimensional datasets like high-resolution medical scans or satellite images.
Tests demonstrate that Cai’s encrypted images are much better at resisting attempt to reverse-engineer them to view the original image than conventional encryption methods.
There is a pressing need for secure, efficient, and scalable encryption methods for a wide range of digital image types. The current research with its novel combination of a six-dimensional CNN and the use of a differential evolutionary algorithm could make those sensitive digital images more secure than ever before.
Cai, Z. (2024) ‘Chaotic colour image encryption based on differential evolutionary deep learning’, Int. J. Information and Communication Technology, Vol. 25, No. 7, pp.57–74.