An Improved Image Multi-Dimensional Chaos Encryption Algorithm Based on CUDA

Due to the intrinsic features of digital images such as bulk data capacity and high redundancy, traditional encryption algorithms designed for textual information are not suitable for image encryption. With their mixing ability, and their sensitivity to the initial conditions and parameters of chaotic maps, chaos-based encryption has suggested a new and efficient way to deal with the intractable problem of highly secure image encryption. However, as the number of calculation steps and pixels being encrypted increases, the encryption time will increase too. Therefore, this study proposes multi-dimensional chaotic encryption, implemented by parallel programming on GPUs using the Compute United Device Architecture (CUDA) to reduce the redundancy of time-consuming processes in encryption. A 2D Arnold cat map was used for pixel position shuffling and 3D Liu chaotic system was used for pixel gray-scale encryption. Experimental simulations have been carried out, and the results prove high efficiency of the proposed algorithms and turn out to be applicable to real-time applications.