[vlc-devel] Ephemeral timeout not working on subpictures

Peter Tap ptrtap at yahoo.com
Tue Apr 10 17:26:20 CEST 2012


The location is generated randomly all over the screen. Each frame shows the watermark on a different location. It will be very hard for the attacker to edit frame by frame.

The random number generator is the one that VLC provides (vlc_lrand48). I can see that the numbers are getting displayed at different location.

What is confusing is why do I see the numbers. The timeout value is for just one frame.

Regards,
Peter



________________________________
 From: Frederic YHUEL <fyhuel at viotech.net>
To: Mailing list for VLC media player developers <vlc-devel at videolan.org> 
Sent: Tuesday, April 10, 2012 1:43 AM
Subject: Re: [vlc-devel] Ephemeral timeout not working on subpictures
 
On Tue, Apr 10, 2012 at 12:57 AM, Peter Tap <ptrtap at yahoo.com> wrote:
> Hello,
>
> I am trying to implement a watermarking scheme that is not visible to human
> eye. The idea is to display a small subpicture at random location for each
> frame. There are 24 frames per second. Human eye should not catch the
> subpicture if it is displayed only for 1/24th part of a second.
>
> My computed timeout for each subpicture is CEIL((1 * 1000 * 1000), 24)
> nanoseconds.
>
> In my video filter, here is the relevant code for subpicture settings:
>
>     p_spu->i_start = date;
>     p_spu->i_stop  = date + p_sys->i_timeout; // timeout from above
> calculations
>     p_spu->b_ephemer = true; // the subpicture is ephemeral
>     p_spu->b_fade = false; // no fading. Just disappear
>
> However, when I run vlc, I still see the subpicture appearing momentarily.
>
> Thinking that my calculation is wrong, I modified the stop time to current
> time + 1 nanosecond.
>
>     p_spu->i_stop  = date + 1;
>
> However, that doesn't seem to make any difference. I still see the image
> momentarily on the screen.
>

Is it possible that you use a really crappy prng? Did you check that
the location actually change each time?

And btw, such a watermarking scheme doesn't seem really robust... I'm
not a specialist, but an attacker could easily remove the watermark,
no?

-- 
Frédéric
_______________________________________________
vlc-devel mailing list
To unsubscribe or modify your subscription options:
http://mailman.videolan.org/listinfo/vlc-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20120410/d07b0474/attachment.html>


More information about the vlc-devel mailing list