Design and development of joint image compression encryption framework using chaotic compressive sensing

Abstract

The proliferation of smart devices and pervasive computing has resulted in the increased use of multimedia data. The deluge of multimedia data leads to a drastic dearth of communication bandwidth, storage resources and the need for secure data transmission. The aforementioned requirements have resulted in the emergence of joint compression-encryption framework for multimedia communication. The objective of these frameworks is to reduce computation complexity, enhance security and speed in communication. Compressive Sensing helps simultaneous compression and encryption of a sparse signal or a compressible signal. A low complex sampling process, which is computationally secure, samples the signal at a rate lesser than the traditional Nyquist criterion. The sampling or the measurement matrix compressively samples the signal and the matrix acts as the secret key of the encryption process. Traditionally random measurement matrices are used as the sampling matrix. However, random matrices are prone to security attacks, the overhead to securely store and transmit these matrices are high. In this research work, chaotic compressive sensing based joint compressionencryption frameworks for images are proposed. Chaotic systems have properties like high sensitivity, long-term unpredictability, ergodicity and random-like behavior. The sensitivity to initial conditions and mixing property of chaotic systems are mapped to confusion and diffusion operations. Also, the chaotic measurement matrix satisfies the restricted isometry property with high probability. The necessary and sufficient condition for perfect recovery in compressive sensing applications is that the signal must be sparse, and the sampling matrix must satisfy restricted isometry property. The sparse representation of the image can be obtained using a fixed or learned basis A learned dictionary and compressive sensing based image compression-encryption is proposed. Sparse coding is used for finding the sparse representation of the images. A Gaussian measurement mat