1. Watermarking for Image Authentication ( Fragile Watermarking )
Multimedia Security

2. Desirable Features
to be able to determine whether an image has been altered or not;
to be able to locate any alteration made on the image;

3. to be able to integrate authentication data with host image rather than as a separate data file ;
the embedded authentication data be invisible under normal viewing conditions;
to allow the watermarked image be stored in lossy-compression format.

4. Block Diagram of Embedding Process
Original image
Lookup Table (LUT)
Entropy Coding
Block DCT
Marked Coeffs.
Quant
Embed
Data to be embedded
Marked JPEG Image
(Watermark is embedded into the quantized DCT coeffs. via a look-up table)

5. 1. Embedding Binary Bits Via Table Look-up
A proprietary look-up table (LUT) is generated before hand by the owner of the image or the digital camera manufactures.
The table maps every possible value of JPEG coeff. randomly to “1” or “0” with the constraint that runs of “1” and “0” are limited in length.

6. To embed a “1” in a coefficient, the coeff
To embed a “1” in a coefficient, the coeff. is unchanged if the entry of the table corresponding to that coeff. is also a “1”.
If the entry of the table is a “0”, than the coeff. is changed to its nearest neighboring values for which the entry is “1”.
The embedding of a “0” is similar.

7. Let denotes the original coeff.
denotes the marked coeff.
is the bit to be embedded in
LUT(.) is the mapping by look-up table :

8. Then, the embedding process can be represented as :if
if , and
The extraction of the signature is simply by :
Where is the extracted bit.

9. Ex : ×  LUT ‥‥‥ -75 -74 -73 -72 ‥‥ 23 24 25 ‥‥‥
‥‥‥ ‥‥ ‥‥‥ ×
-73 1
Embed “0” (changed)
-74 36
-73 8 … 24 5
original
embedded
-74 24
Marked 24 
Embed “0” (unchanged) 24

10. In this example, zeros are embedded into two AC coefficients with values “-73” and “24” of an 8×8 image block. The entry in LUT for coeff-value “-73” is “1”. So to embed a “0”, we go to its closest neighbor for which the entry is “0”.
In this case, “-73” is changed to “-74”.
Since the entry for coeff-value “24” is “0”, it is unchanged.

11. 2. Consideration for Minimizing Distortion
To ensure the markings are invisible :
The run of “1” and “0” entries in the LUT is constrained to avoid excessive modification on the coeffs.
DC coeff. in each block is not changed to avoid blocky effect.
Small value coeffs are not modified to avoid high freq. distortion.

12. The above constraints make the number of bits that can be embedded varies significantly from block to block. Also extraction error may occur due to image format conversion and other causes.
For these reasons, we choose to embed “one bit per 8×8 block” by embedding the same bit to all embeddable coeffs in the block and detecting that bit by “majority voting”.

13. Remarks
The detection of an error in the embedded bit signals alteration of that block.
The constraint of leaving DC coeffs. and small valued AC coeffs unchanged makes it impossible to embed data in smooth blocks where all AC coeffs are very small and hence unembeddable.

14. 3. Choice of Embedded Data for Authentication
The authentication data we embed in an image consists of a “visually meaningful binary pattern” and some “content features”.
The visually meaningful pattern serves as a quick check for alteration and location of possible alteration.

15. A simple example of content feature is the “most significant bit” of macroblock mean intensity.
Another example is the “sign” of intensity difference between macroblocks.
The combination of these two types data is suitable for such applications as image authentication for “trustworthy” digital camera.

16. block Quant Extract visualize DCT Embedded Features compare Features
Original
Image LUT
Marked
block Quant Embed coeff. Entropy Compressed
DCT coding Marked
Image
features Binary
pattern
Test Image
(decompressed) LUT Binary Pattern
block Quant Extract visualize
DCT Embedded Features compare
Features
Embedding and Extraction Procedures

17. 4. Smooth Region Embedding
Authentication data cannot be embedded in smooth blocks because the AC coeffs are small.
This shortage can be solved by embedding data in blocks that are not smooth but whose location bears a “fixed relationship” to the smooth block in question ——backup embedding.

18. 5. Color Image and Video
For color images, one may work in YCrCb coordinates and use the proposed approach to mark luminance components while leaving chrominance components unchanged ; or one may apply the approach to chrominance components as well as to embed more data. One may also work in other color coordinates, such as RGB.
For MPEG-video, one may mark the I-frames using the proposed scheme. The I-frame serial number can be used as part of embedded data to detect modification such as frame reordering and frame dropping.

19. References
[1] Yeung & Mintzer, “Invisible Watermarking for Image Verification”, J. of Electronic Imaging, pp , July 1998.
[2] P.W.Wang, “A public Key Watermark for Image Verification and Authentication, “ IEEE Inte’l conf. Image Processing, pp , Oct
[3] Wu & Liu, “Watermarking for Image Authentication,” IEEE Inte’l conf. Image Processing, pp , Oct
[4] Xie & Arce, “ Joint Wavelet Compression and Authentication Watermarking,” IEEE Inte’l conf. Image Processing, pp , Oct