Steganography using overlapping codebook partition

1. Steganography using overlapping codebook partition Source: Signal Processing,Available online 28 November 2007 Authors: Yung-Kuei Chiang, Piyu Tsai Report: Jiun-Lwen Liang Date: 2012/06/07

2. Outline • Introduction • The proposed scheme 1. codebook partitioning 2. embedding procedure 3. extracting procedure • Experimental results • Conclusions

3. Introductions • Communications security over the Internet is becoming more and more critical because of the digitization of data and networking of communications. • The steganography using overlapping codebook partition is proposed to extend the hiding capacity of a cover image and preserve the quality of a stegoimage.

4. Introductions • Jo and Kim proposed an index-based watermarking scheme. • Chang and Wu presented an adaptive VQ-based data hiding scheme which is based on a codeword grouping technique.

5. Related work - VQ • Vector quantization (VQ) – Codebook generation 0 1 2 . . . N-1 Training Images Codebook

6. Related work - VQ • Vector quantization (VQ) – Encoding Original Image Index table Codebook

7. Related work - VQ • Vector quantization (VQ) – Decoding VQCoded Image Index table Codebook

8. The proposed scheme Codebook partition

9. sort the SEDs of each codeword between all other codewords from small to big clustered into four-member sub-cluster ( summed SED < TH1) clustered into three-member sub-cluster ( summed SED < TH2) clustered into two-member sub-cluster ( summed SED < TH3) clustered into single-member sub-cluster The proposed scheme -codebook partitioning

10. Codebook Codewords 0 1 (127, 32, …, 56) 2 3 … 254 255 The proposed scheme -codebook partitioning Codebook partition algorithm Squared Euclidean Distance (SED) Thresholds TH1, TH2, TH3

11. The proposed scheme -codebook partitioning TH1=455 0+80+90+100=270 0+80+110+120=310 0+110+160+190=460 0+120+220+230=570 0+180+185+190=555 0+90+140+160=390 0+100+140+210=450 0+270+310+320=900 60 → 60,61,65,66 61 → 60,61,62,63 65 → 60,62,65,66 66 → 60,61,65,66 TH2=350 TH3=200 Residual 0+110+160=270 0+120+220=340 0+180+190=370 0+270+310=580 67 →67 0+190=190 0+270=270 62 → 62,61,65 63 → 63,61,64 64→64,62

12. The proposed scheme -embedding procedure (four-member sub-cluster) index 209 199 secrete message: 0 1 1 0 1 1 0 0 1 1 … Index table

13. The proposed scheme -embedding procedure (three-member sub-cluster) Secret bits valued at 5 embed bits : 101 71 67 Index table

14. The proposed scheme -embedding procedure secrete bit: 01 Find the most similar one to the original index 175

15. The proposed scheme -Extracting procedure index … 199 Index table … … 209 secrete bit : 01 codebook

16. The proposed scheme -Extracting procedure Index table codebook Secret bits 101

17. (f) Gold (1) IEEE (2) CCU (3) NUU Experimental results 512X512 pixels 128X128 pixels

18. Experimental results(2/5) Original image Embedded image “IEEE” “CCU” “NUU”

19. Experimental results

20. Experimental results Cover image (512×512) Secret data (256×256= 65536 bits)

21. Experimental results Secret data 256x256

22. Experimental results

23. Conclusion • This method increases the hiding capacity and still have the great image quality.