ESSENCE: GPU-based and dynamic key-dependent efficient stream cipher for multimedia contents

Abstract

Data Confidentiality (DC) is considered one of the most important security services. Currently, a set of existing cipher algorithms is being used to ensure DC. However, researchers constantly investigate the design and implementation of more efficient cipher schemes. To this end, different versions of AES have been implemented efficiently on GPUs to increase the efficiency over big data. However, AES implementation on GPU exhibits limitations in terms of latency and hence, it might not be a suitable solution for high data rates in modern systems and applications. This often leads to a trade-off between system performance and security level. To address these challenges, we propose “ESSENCE”, a lightweight stream cipher scheme, which combines two different Pseudo-Random Number Generators (PRNG), and based on a dynamic key approach. The scheme achieves a high level of security with minimal latency and required resources when compared to existing cipher standards such as AES. Moreover, the implementation of the proposed dynamic key-dependent cipher scheme on GPU is more efficient compared to all existing AES implementations on GPUs. Experimental results indicate that the proposed cipher is highly efficient with a throughput more than 115 GB/s on a Titan X GPU, and more than 372 GB/s on a Titan V100 GPU. Thus, ESSENCE can be considered as a promising stream cipher candidate with high randomness degree (BigCrush of TestU01), periodicity, and key sensitivity.