Re: [WM]: On Attack by wmking wmked images(I.J.Cox 97')

["Corina Nafornita" <corina.nafornita@etc.upt.ro>]

Hi,
You have to look at the problem from a different perspective I think. The multiple watermarks are embedded in the same location
(i.e., the key from the original image), for different purposes: say one identifies the owner, the other the creator of the work and
so on.

Or perhaps an attacker simply wants to destroy the first watermark by inserting his own into the same location.
Look at the phrase: "this attack is equivalent to adding noise to the frequency bins containing the watermark".

Then you get all the five peaks, because what you are doing is trying to measure the similarity between W1, W2, W3, W4, W5 and the
extracted watermark which would be more or less W1+W2+W3+W4+W5, intuitively speaking.

Obviously if the watermarks are not correlated, you will have 5 peaks.

What you are trying to do is also ok, but it is not what the authors claim in their paper. You only get one peak, because you only
"look" for the first watermark, in the locations where it was embedded. I hope I make myself clear. Please correct me if I am wrong.

I suggest you read A DCT-domain system for robust image watermarking , Barni, Bartolini, Cappellini, Piva, section 4.8 (same
attack).

Good luck,
Corina

> hi,all:
> i got some trouble when implementing the watemarking system proposed 
> in I.J.Cox 's paper:
>
>
>  I. Cox, J. Kilian, F. T. Leighton, and T. Shamoon, *"Secure*
>
> *spread spectrum watermarking for multimedia,"* *IEEE Trans.*
>
> *Image Processing, *vol. 6, pp. 1673=A8C1687, Dec. 1997.
>
>
>
> On Attack by watermarking watermarked images,i cannt get proper 
> results = as the paper claimed.
>
> the watermark detector response show exactly one peak.
>
>
>
> my matlab code section is as follow:
>
> ----------------------------------------------------------------------
> ---=
> ------
>
> *%Multi-watermarking*
> disp('Attack by watermarking watermarked image...');
>
> n=3D1000;
> alpha=3D.1;
> cnt=3D5;
>
> for k=3D4:6
> i=3Dimread(file{k});
> i=3Dim2double(i);
>
> i2=3Di;
> for index=3D1:cnt
>     w{index}=3Drandn(1,n);
>     i2=3Dcox_embedwmk(i2,w{index},alpha);
> end
>
> w2=3Dcox_extractwmk(i,i2,n,alpha);
>
> *%SNR-PSNR-SIM Calculation*
> if ndims(i)>2
>     i_YIQ=3Drgb2ntsc(i);
>     i2_YIQ=3Drgbn*
> if ndims(i)>2ntsc(i2);
>     snr=3DSNR(i_YIQ(:,:,1),i2_YIQ(:,:,1));
>     psnr=3DPSNR(i_YIQ(:,:,1),i2_YIQ(:,:,1));
> else
>     snr=3DSNR(i,i2);
>     psnr=3DPSNR(i,i2);
> end
>
> disp(sprintf('SNR=3D%4.2f,PSNR=3D%4.2f',...
>     snr,psnr));
>
> figure;
> subplot(121),imshow(i),title('original image'); 
> subplot(122),imshow(i2),title('multi-wmked image(5 times)');
>
> *%uniqueness of watermark*
> j=3D1;
> for index=3D1:1000
>     if mod(index,100)=3D=3D0 && j<=3Dcnt
>        w_original=3Dw{j};
>        j=3Dj+1;
>     else
>         w_original=3Drandn(1,n);
>     end
>     sim_test2(1,index)=3Dcox_sim(w_original,w2);
> end
> figure,stem(sim_test2,'marker','none');
> title('Uniqueness of watermark');
>
> end
> ----------------------------------------------------------------------
> ---=
> -------------------------------------------
>  can anyone tell me what is wrong there?
> thanks in advance ^_^
>
> --
> Gao Zhengwei
> State Key Lab of CAD/CG,ZheJiang University gaozhengwei@google.com


-- 
Corina Nafornita
Politehnica University of Timisoara
Communications Dept.
300223 Timisoara, Romania

e-mail: corina.nafornita [at] etc [dot] upt [dot] ro
phone: +40 256 403 308
mobile: +40 722 782 359


______________________________________________________________________________

Watermarking Mailing List - http://www.watermarkingworld.org/ml.html
To unsubscribe send email to "majordomo@watermarkingworld.org" with
"unsubscribe watermarking YOURMAIL" in the body.
______________________________________________________________________________