FPGA-based Hardware Acceleration for Image Copyright Protection Syetem Based on Blockchain

Abstract

The number of images published on the Internet has increased exponentially. Image attribution and content tempering happen when users easily get and edit the images. Blockchain as an emerging technology, has a good potential in copyright protection due to its transparency and immutability. In this paper, we propose an image copyright protection system based on the fusion of deep neuron network and blockchain to address the image certification issue and identify edited region of the original image. Deep neuron network is used to extract features of the original images, which are then stored in the blockchain with owner’s information. Blockchain is used as a secure database to store important information to provide time and place of existence of work. InterPlanetary File System (IPFS) is used to store large amount image data to achieve decentralization which integrate with blockchain well. This system can not only support image copyright registration and protection, but also detect edited regions to some extent. To accelerate two most time-consuming operations in the systems, i.e. DNN and SHA-256, a novel on-chip computing architecture is needed. We design a scalable DNN accelerator and SHA-256 using field-programmable gate array (FPGA). The DNN kernel can adapt to different network topologies. We also implement SHA-256 function on FPGA. Experimental results show that the whole acceleration system can achieve up to 40x speedup comparing to software implementation on CPU.