淺談 浮水印技術

1. 淺談浮水印技術 94321026 資工三 呂若堇

2. Outline • Purpose and Application • Categories • Requirement • System framework • Watermark Example • Conclusion and Future work

3. Purpose and Application • Purpose: • To protect intellectual property rights • Use the weakness that human eyes can’t detect slight changes between the images • After going through en/decryption, compression or other image processing, the information embedded can still be extracted or detected by algorithm

4. Purpose and Application (cont.) • Application: • Broad Monitoring • Transaction Tracking • Copy Control • Device Control • Contentauthentication • Fingerprinting

5. Categories • Watermarked image I = I + W • Visually: • Visible • Invisible • Technically: • Robust watermark • Fragile watermark • Semi-fragile watermark

6. Categories (cont.) • Embedding--technically: • Spatial domain • Frequency domain • DCT, Discrete Cosine Transform • DFT, Discrete Fourier Transform • DWT, Discrete Wavelet Transform • Hybrid • Visual Cryptography

7. Requirement • Transparency: • PSNR（Peak Signal to Noise Ratio） • If PSNR > 30dB, means that human eyes can’t distinguish the difference between I and I’ • Robustness: • After several attacks, we can still extract the hiding information

8. Requirement (cont.) • Security: • Emphasize on specific region of original image, in that adversary find out the watermark through those specific regions • Do not allow the adversary to determine the location of the mark in the watermarked。 • Oblivious: • No need for host media to help extract the watermark

9. Requirement (cont.) • Multiple Watermarking: • When embedding another watermark using the same mechanism, it won’t affect afore-watermark • Still a challenge!! • Unambiguous: • The extracted logo must be specific • NC（Normalized Correlation）, if both are the same, NC=1, otherwise, NC decrease

10. System framework • Three main issues for designing it: • 1. Design a watermark signal W to be embedded to the host media • 2. Design an embedding algorithm • 3. Design corresponding extraction method

11. System framework (cont.) original image (host signal) Secrete key Crop, scale, Rotation, Translation, … Smooth, Meadian filter, Gaussian noise, … JPEG, SPIHT EZW … Secrete key Extracting Algorithm Embedding Algorithm watermarked image (embedded signal) Attack

12. Watermark Example • DWT: • Haar函數小波轉換: Most significant part

13. Watermark Example (cont.) • Watermark embedding: Binary Watermark ,W Host image, I Torus Automorphism ,TA Embedder Wavelet transform, T Gray scale Inverse Wavelet Transform, T’ Image “Pepper” with hiding image “Jet” Watermarked image, I’

14. Watermark Example (cont.) • Watermark detection: Attacked image Host image Wavelet transform, T Detector Wavelet transform, T Inverse TA Reconstructed Watermark, w’

15. Conclusion& Future work • Future work: • Deadlock Problem: • When it comes to multiple watermark, mechanism nowadays can’t make sure which one is authentic • Focus of watermark technique is no longer on how to embed and the performance of robustness

16. Conclusion& Future work • Conclusion: • Video: • I-Frame, motion vector • Must stand changes in frame rate, recording to tape • Audio: • Slightly alter signal amplitude on time domain • Must stand temporal filtering, recording on audio tape

17. 參考資料 • http://en.wikipedia.org/wiki/Digital_watermarking • http://www.usb20.idv.tw/usb20/viewthread.php?tid=75 • http://wmail.hcu.edu.tw/~cchen34/20071018_01.ppt • http://211.79.59.247/Course/Infosecurity20040507/slide/Watermark_and_Authentication_2004_0502.ppt