HLS Scheduling Driven Encoded Watermarking for Secure Convolutional Layer IP Design in CNN

This paper presents a novel high level synthesis (HLS) scheduling driven robust watermarking methodology for convolutional layer intellectual property (IP) design in convolutional neural network (CNN), useful for consumer electronics (CE) systems. The results of proposed approach yielded enhanced security in terms of tamper tolerance stronger entropy than prior works, at nominal design/latency overhead.