An upgraded image encryption based on nonlinearity concept in blind source separation

Blind source separation is widely used in image and speech encryption in the last two decades. The main drawback of BSS-based cryptosystem is its linear structure, which makes it vulnerable against all kinds of cryptanalytic attacks. Inserting the nonlinearity in encryption phase is a solution to make the cryptosystem more secure. In this paper, we focus on the nonlinearity effect on BSS-based cryptosystem. To increase the security, four types of nonlinear functions are introduced and evaluated. The presented nonlinear functions increase the nonlinearity effect of cryptosystem. They improve the security by 32-45% in different kinds of sparse and none sparse plain images based on FSIM criteria. The FSIM is a similarity measurement criterion based on human visual system (HVS) established on images feature. Due to the successful operation of the introduced nonlinear functions, a new cryptosystem is proposed based on post-nonlinear BSS. Compared with previous presented BSS-based cryptosystems, which often were vulnerable against cipher text-only differential attack, the proposed cryptosystem is robust against this attack and has a high sensitivity to key. Furthermore, due to the unconditional decryption process, there is no restriction on the sparse and correlated plain images for encryption and decryption. Simulation and analysis display the proposed method increases the security of BSS-based cryptosystem.