Evaluation Criteria for Visual Cryptography Schemes via Neural Networks
Yunchao Wang, Yunfa Li, Xiao‐Nan Lu · 2020
Visual cryptography schemes are developed for image security, where encryption is realized by distributing a secret image into shares and decryption is done only by stacking the shares. Human eyesight is usually used to evaluate the security and performance of visual cryptography schemes. However, objective criteria for visual cryptography schemes are not yet established. In this paper, by the aid of neural networks, we propose two criteria called encryption-inconsistency and decryption-consistency for evaluating the shares and the recovered images, respectively. We also implemented the experiments for two representatives of visual cryptography schemes by applying three popular convolutional neural networks (CNN) to adopt our proposed criteria.