Watermarking/Digital signature tool...

[fidels]

My boss is releasing a short movie (~15 min) in a few days, and wants to digitally watermark the DVDs he'll be sending out to the festivals, in order to be able to know who leaked the movie if it appears online (which was what happened last year).

I've been searching for a decent tool to acomplish this for a while now (both in this forum and in the web in general), and the only program I found nearing what I need is MSU's StegoVideo. However, my tests show that information stored with this plugin will only survive very few video modifications (and certainly not those that I've seen in pirated movies): it does well when levels, brightness or smoothing is applied, but fails to get through when cropping or resizing comes into play. Also, it doesn't survive successive encodings with Xvid, or combinations of a couple or more filters...

Therefore, my questions are: Is there any other tool to hide some information (in this case, a serial number) in a video stream? And if not, is there any way to improve StegoVideo's performance? Or any other method to determine which copy of a video gets leaked (from about 20 possible copies)?

I know the big studios use digital watermarks for their screeners, but I'm guessing they keep their filters in a safe, and never release them to the public. However, if anyone knows about the methods they use to conceal information, please tell me, because then maybe I could write my own plugin...

Anyways, thanks in advance for your help!

[Dark Shikari]

Another problem with this idea is if one has two copies of the video, most stenography techniques can be broken unless they were specifically designed to survive recombination.

[kumi]

Have you looked into tools for watermarking the audio?

[fidels]

@Dark Shikari - Well, I'm hoping that if there is a leak, it will come from a single source. If not, I guess I'm out of luck :-S

@kumi - I have, and for what I've seen, information tends to get lost even more when stored in the audio stream. I guess this is due to the fact that most of the time the audio is compressed far more than the video. Anyways, I have already decided to watermark the audio as well as the video, as some sort of extra redundancy check...

[scharfis_brain]

what about just overlay a text over the video with a deviation of a amount of one luminance step.

This is almost invisible and will survive multiple recodings and heavy filtering.

One can recover the overlaid text by averaging all frames and incresing the contrast afterwards.

of course, if a ripper really looks for this kind of near-to-invisible watermarking he will notice it and furthermore is able to remove it.

[foxyshadis]

I'd think watermarking the audio would be much faster and not much more complicated, given how easy it is to change the offset (which recompression won't effect). If you're looking at rolling your own, twenty thousand samples with 2-4 thousand sample buffers to smooth transitions between them would be inaudible, and you can decided how much tolerance you want each to have (thus how many bits you can have in each). I don't know why audio watermarking tools don't do this.

The decoder would just need to average 20 thousand sample windows until it finds a recognizable start code of some sort.

[scharfis_brain]

also one could change the content of the movie slightly.

like randomly deleting a frame before or after a scene change.

[Shinigami-Sama]

you could overlay a random image with a 99.9% alpha every X frames
a parser could still see it, but a person wouldn't be able to

[olyteddy]

Free watermarking tool today only: http://www.giveawayoftheday.com/

[scharfis_brain]

Everthing the tool offers can be done using avisynth...

[Shinigami-Sama]

Everthing the tool offers can be done using avisynth...

TBH
avisynth can almost make your coffee in the morning, if you want to spend asking it how to

[scharfis_brain]

nah, first I thought it could do invisible watermarking like digital signatures.
But after I installed it I was disappointed, cause it just is a more or less comfortable logo-burning program.

[venkat.siddareddy]

you could overlay a random image with a 99.9% alpha every X frames
a parser could still see it, but a person wouldn't be able to


Hi,
can some one tell me how to overlay a random image with 99.9% alpha every 100 frames.

I am planning to use this for invisible watermarking.

Also please help me to check for the watermark after it has been recorded using a video cam or screen capture


Thanks
Venkat

[LoRd_MuldeR]

Hi,
can some one tell me how to overlay a random image with 99.9% alpha every 100 frames.

I am planning to use this for invisible watermarking.

Also please help me to check for the watermark after it has been recorded using a video cam or screen capture

It seems what you want is a "true" cryptographic/steganographic digital image/video watermark. I doubt you'd be able to achieve this with a simple "overlay" image ;)

[unix_sansei]

Hi,
can some one tell me how to overlay a random image with 99.9% alpha every 100 frames.

I am planning to use this for invisible watermarking.

Also please help me to check for the watermark after it has been recorded using a video cam or screen capture


Thanks
Venkat

a random image is just noise. so you are artificially increasing the psnr for one frame? if you bump the psnr too much any decent image processing software can easily pick up on that without even seeing the image. no natural phenomena have such an extreme discontinuity in the noise floor and if the spike is the psnr is every X frames then all i would have to do it fft the alpha/gamma/chroma channel to find the frequency. you would have to alter not just the image but the frame's timing as well. you would have to sync your random seed generator whatever your viewer/detector is.

it seems like you could use gmagick or imagick hook in a random number generators to it. maybe even avisynth.

[venkat.siddareddy]

a random image is just noise. so you are artificially increasing the psnr for one frame? if you bump the psnr too much any decent image processing software can easily pick up on that without even seeing the image. no natural phenomena have such an extreme discontinuity in the noise floor and if the spike is the psnr is every X frames then all i would have to do it fft the alpha/gamma/chroma channel to find the frequency. you would have to alter not just the image but the frame's timing as well. you would have to sync your random seed generator whatever your viewer/detector is.

it seems like you could use gmagick or imagick hook in a random number generators to it. maybe even avisynth.

Hi,
Thanks for your answer but I am not sure if I understand it fully.
Can you please give an example of the same using avisynth.

[ninjashoes]

Free watermarking tool today only: http://www.giveawayoftheday.com/

This is useful, thanks