ABSTRACT

With the proliferation of the internet of things (IoT) and device-to-device (D2D) communications enabled by the boom of mobile computing technology, secure high-speed communication has now become indispensable in our daily life. This is especially true when potentially private data are being continually sensed by and communicated among mobile devices as they exist in a world of interconnected inanimate objects, which is also one of the main themes of the upcoming 5G revolution. As the amount of data to be secured for high-speed communications is vast, there is a need to ensure that the block ciphers used for encryption are deployed without incurring significant computational cost. In this paper, we present fast implementations of recent industry standard block ciphers in typical embedded platforms, consisting of multi-core CPU (ARM A15 and A7) and GPU (Mali T628). We implemented the conventional block cipher (AES) and lightweight block ciphers (CLEFIA, SIMON, SPECK and PRESENT) optimized for fast computation. We also analyze the energy efficiency of these block ciphers computation in CPU and GPU, as low power consumption is crucial for the embedded system. Our experimental results show that the embedded GPU is not only able to compute block ciphers faster than conventional CPU but also consumes significantly less power.

摘要

随着移动计算技术的蓬勃发展，物联网 (IoT) 和设备对设备 (D2D) 通信的普及，安全的高速通信现在已成为我们日常生活中不可或缺的一部分。当潜在的私人数据存在于一个相互连接的无生命物体的世界中时，移动设备不断地感知并在它们之间进行通信时尤其如此，这也是即将到来的 5G 革命的主题之一。由于高速通信需要保护的数据量很大，因此需要确保在不产生大量计算成本的情况下部署用于加密的分组密码。在本文中，我们展示了在典型嵌入式平台中最新行业标准分组密码的快速实现，由多核 CPU（ARM A15 和 A7）和 GPU（Mali T628）组成。我们实施了针对快速计算进行优化的传统分组密码 (AES) 和轻量级分组密码（CLEFIA、SIMON、SPECK 和 PRESENT）。我们还分析了 CPU 和 GPU 中这些分组密码计算的能效，因为低功耗对于嵌入式系统至关重要。我们的实验结果表明，嵌入式 GPU 不仅能够比传统 CPU 更快地计算分组密码，而且功耗显着降低。