Synchronization-Based Spatiotemporal Deterministic Random Stream Cipher And Key Exchange Protocol

Owing to its sensitivity to initial conditions and ergodicity seems to be a model for the classic Shannon requirements of confusion and diffusion, chaos has potential application in secure communications. Unfortunately, lots of negative cryptanalysis results have shown that the security of chaos-based communications is rather low. A new nonlinear dynamic phenomenon named deterministic randomness, which bridges chaos and randomness, extends the theory of dissipation and conservative dynamics, and may play a key role in increasing the security of chaotic ciphers. This dissertation mainly focuses on the self-synchronization stream cipher based on spatiotemporal deterministic randomness and neural cryptography.Firstly, the conventional cryptography, synchronization-based chaotic cryptography and neural cryptography are briefly introduced.Secondly, we investigate the highly efficient key exchange protocol based on the mutual learning of neural network. After detailed analyses of flipping attack and cooperating attack strategies, a novel neural cryptography scheme with divided synchronization and confusion unit is proposed based on the explainable model for asymptotic deterministic randomness. The synchrony and security is unified in this framework that the hidden output is confused by the non-reversed nonlinear transformation while there still exists necessary driving force to ensure the synchronization. Both analytical and experimental results show that the success probability of flipping can be decreased to the level of brute force attack, and that of cooperating attack still exponentially decreases with the range of the weights without any influence on the synchronization. Furthermore, the security of the proposed scheme is also analyzed under an advanced flipping attack.Thirdly, a class of spatiotemporal chaos self-synchronization based stream cipher system stated as the strongest scheme among the chaos secure communication systems is evaluated. The prototypical system is roughly equivalent to an asymptotic deterministic random system. However, three weaknesses in security and encryption efficiency including the very limited confusion and diffusion effect of single lattice, very strong coupling strength for parallel encryption, and degeneration to one-dimensional under constant driving are pointed out. Based on the weakness that the high-dimensional chaotic system will degenerate to be one-dimensional under constant driving and will be no more sensitive to the slight perturbation on the specially designed driving, two types of differential-like chosen cipher attack schemes are suggested, and then one can reveal the underlying keys within only hundreds of iterations.Finally, we propose a spatiotemporally asymptotic deterministic randomness based self-synchronization stream cipher system. Both analytical and experimental results show that the sensitivity of the proposed scheme with coupling size just three can be equivalent to that of normal spatiotemporal chaotic system with coupling size 280. The parallel encryption can be possible in one-dimensional structure because all the outputs from different sites are statistically unrelated in wide range of coupling strength. The proposed system, which is an extension of dissipation and conservative dynamical system, can resist the differential-like attack because even under constant driving, the degenerated system is still high dimensional.