RE: [WM]: Query

["Bloom Jeffrey" <Jeffrey.Bloom@thomson.net>]

Vidy, 

If I understand your comment, you believe that attacked content is not valuable if it is visibly degraded from legitimate copies.  I
do not agree.

When embedding a watermark, there are two reasons to maintain very high fidelity.  First, the goal here is usually to make the
watermark invisible or unobtrusive.  Second, if the mark is visible, it may be easier for an adversary to remove it.

However, an attacker has different constraints.  An attacker can usually introduce visible artifacts in a removal attempt.  Of
course this depends on the application, but look at the movies available on a p2p network. Horrible quality, yet still useful and
still (apparently) able to impact legitimate sales.  In fact, I think that the mpeg compression used for digital broadcasts often
introduces unacceptable visual artifacts, yet this degradation does not appear to lessen its value.

Given a fidelity measure, we can specify these constraints quantitatively.   There is a fidelity threshold controlling the
embedding.  All embedder output should be above this threshold.  This is, of course, dependent on the application.  Similarly, there
is a fidelity threshold below which we cannot guarantee reliable detection.  This applies to both robustness and security against
unauthorized removal.  Typically this second fidelity threshold is much lower than the embedding fidelity threshold.

-----------------------------------------------------------------
Jeffrey A Bloom, Content Security Group
Thomson Multimedia Corporate Research, Princeton, NJ, USA
(609) 987-7727, (609) 987-7299 fax
http://www.geocities.com/Jeffrey_Bloom

> -----Original Message-----
> Vidyasagar Potdar <Vidyasagar.Potdar@cbs.curtin.edu.au> wrote:
> Hello
> I have a query regarding watermark attacks.
> 
> My assumption is that if watermarked images are attacked, and the
>watermark is successfully removed the quality of the image should  not
>be changed.
> 
> If the quality of the image is changed then there is no need for an
>adversary to attack the images because there is no use for  having
>those modified images.
> 
> Specaially i refer to some attacks that have been listed on some of
the
> benchmarking applications like Checkmark etc.
> For example Threshhold, Bending, Warping, Arbitrary Linear Transform,
> Dithering these attacks change the quality of the image to such an
> extent that it could be any useful (my assumption may b i m
wrong)
> 
> Can anyone comment on these views????
> Does a watermarking algorithm really needs to be robust against these
>attack from a practical point of view???
> 
> Thanks very much
> 
> Regards
> 
> Vidy
> 
> 


______________________________________________________________________________

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