Development of Efficient Frame(s) Encryption Techniques

Abstract

The protection of personal information is of utmost significance in today’s digitally connected society. In this digital era, encryption is seen as a rapidly developing trend. Despite the development of sophisticated modern information security techniques, storage, and retrieval procedures, the internet’s widespread use, unsecured transmission channels, and hacking tools continue to pose significant issues. To address the issue of privacy and data security, several techniques have been developed, yet there is still need of developing efficient and secure cryptosystem for images/videos. In this thesis, several time efficient and secure image encryption techniques are proposed for different applications. We propose a hybrid chaotic map which is a combination of two 1-D chaotic maps, i.e Gauss and Square. These two maps have simple structures, however, by combining these nonlinear chaotic maps results in highly complex and unpredictable chaotic map. Using this chaotic map, we further propose an efficient chaos-based encryption technique for images/videos. The chaos based image encryption technique employs chaotic map proposed above in scrambling and substitution steps to encrypt images providing high security. Chaos-based algorithms are useful as they provide images with high security and suitable for encrypting images in real-time applications. Proposed image encryption also provides security in content based image retrieval by efficiently retrieving similar images in encrypted domain. One of the prime utility of this technique may be seen in maintaining privacy of health records of patients on cloud server. This thesis also explores selective encryption technique in which only some part of the image is important and needs complex encryption algorithm to provide high end security. Then a framework is proposed for remote monitoring of vital signs by integrating digital twins technology to enable real-time monitoring with minimal cost overhead. To secure the data stored in third party cloud, all the mathematical computations are performed on encrypted data. We have also proposed a trusted federated learning approach for applications where all the data (in the form of images) can not be stored in central location due to some privacy or legal restrictions. By introducing trust and security in federated learning, activities of malicious users are minimized which leads to more accurately trained model. The proposed techniques are evaluated using various security measures and results are compared with the state of art existing technique to verify the significance of proposed image encryption technique. Simulation results demonstrate that all the proposed techniques exhibit efficiency and produce highly secured images.