类神经网路於视觉密码学之应用.ppt

類神經網路於視覺密碼學之應用 Reporter : 劉益成 Advisor : 虞台文 Contents Overview The Surveys Neural Network Model Visual Cryptography Using Neural Networks Histogram Allocation Problems Conclusion and Future Works Overview What is Visual Cryptography? To achieve visual secret sharing. Shares Binary image printed in transparencies. Target Image --- Message concealment Decryptor --- Eyes Example Applications Key management Message concealment Authorization Identification Entertainment Our Approach Using the Q’tron NN Model. The Surveys on Visual  Cryptography The Evolution Visual Cryptography(1994): Monni Naor and Adi Shamir. Extension (1998): Giuseppe Ateniese, Carlo Blundo, Alfredo De Santis and Douglas R. Stinson. Concept Naor and Shamir (2,2) Concept Naor and Shamir Disadvantages Pixel Expansion --- image size enlarged Participants are all binary images? How to find code books for complex access scheme, e.g., (k,n )? Very difficult Is an access scheme realizable? ? Is a code book existent? Giuseppe Ateniese, Carlo Blundo, Alfredo De Santis and Douglas R. Stinson. Neural Network Model Q’tron---Q-State Neuron Q’tron’s Input State Updating Rule: Q’tron Neural Network System Energy Visual Cryptography Using Qtron Neural Networks Main Concept (2, 2) Access Scheme Image Halftoning Image Halftoning Stacking Rule Satisfaction Stacking Rule Satisfaction Experimental Results Experimental Results Experimental Results Experimental Results Experimental Results Conclusion and Future Works Conclusion Target images and share images use graytone images. Share images size= target images size. Code book is free. Future Works Extend to color images. Neural chip Complete theories for access schemes. 若我們希望,share1的平均gray level值驅於k1,而share2的平均gray level值驅於k2,因此可寫成如下一張的energy finction 為簡化,且share重疊就可decode,因此在此略過 因此,total energy就是分別由halftoning 及stacking rile和使之驅於plain image之三部份所組成 並且對應到tow-dimensional Q‘tron NN 手首先讓我們來討論一個簡單的問題, 如兩張太亮的投影片,不可能合成,全黑, 只有較暗,才有可能 同理,我們在作visual cr