5 ms·

Real terrorists have already been using that for a long time. (For instance, remember those media blurbs about lots of porn being found in the raid on bin Laden

Real terrorists have already been using that for a long time. (For instance, remember those media blurbs about lots of porn being found in the raid on bin Laden?) As such I'd bet they have infrastructure in place to detect such communications. But in the same half-joking vein, some other problems: 1. I recall reading that such stegonagraphically watermarked videos were pretty easy to detect using simple statistical analysis of the noise in the frames. With dedicated hardware for multimedia decoding cheaply and ubiquitously available, frames can be efficiently extracted and the additional number of bytes does little to slow them down. If the hidden data is encrypted, well, you could just do that without the porn. 2. Unless you make it as easy as tapping a key to start communicating, it still does not remove a main barrier to widespread adoption of encryption, which is inconvenience. 3. The problem with porn is, people will just end up paying more attention to the medium rather than the message.

nwh 13y ago
> you could just do that without the porn. I think the idea is that it's something people naturally keep hidden, so it doesn't look suspicious to have handed to someone n a USB drive or sent privately. There's not really much suspicious about people hiding porn.
ippisl 13y ago
I read some research about provably secure steganography , which claim that if you know the statistics of the communication channel, you can build secure steno, at least to the level one way functions. So it is interesting if in reality there is a stego detection tool. www.cs.umn.edu/~hopper/tc-stego.pdf Btw one of the authors is Luis von ahn, the founder of duelingo, the language teaching/translation app.
- im3w1l 13y ago
  > if you know the statistics of the communication channel, you can build secure steno That's a pretty big if though. I mean, if you miss one nuance the NSA didn't, then your scheme could be toast. At least in the high data limit.
  - ippisl 13y ago
    You're right, that's a very big if. Also, most of the research in the field has been done on black and white image, so it not directly usefull. And come to think on it, if we're talking about image stego, what prevents the NSA from controlling the statistical distributions of output of cameras in such a way that stego will be noticed(even after some image transforms) ?
    - IsThisObvious 13y ago
      Maybe the answer is to have a program that adds low scale random noise to as many images as it can before upload to the internet. It would be easy enough to make a one-click batch process, and would help provide cover noise for real communication channels.
- zeckalpha 13y ago
  Also, this group invented the CAPTCHA.

nwh 13y ago

> you could just do that without the porn. I think the idea is that it's something people naturally keep hidden, so it doesn't look suspicious to have handed to someone n a USB drive or sent privately. There's not really much suspicious about people hiding porn.

ippisl 13y ago

I read some research about provably secure steganography , which claim that if you know the statistics of the communication channel, you can build secure steno, at least to the level one way functions. So it is interesting if in reality there is a stego detection tool. www.cs.umn.edu/~hopper/tc-stego.pdf Btw one of the authors is Luis von ahn, the founder of duelingo, the language teaching/translation app.

im3w1l 13y ago
> if you know the statistics of the communication channel, you can build secure steno That's a pretty big if though. I mean, if you miss one nuance the NSA didn't, then your scheme could be toast. At least in the high data limit.
- ippisl 13y ago
  You're right, that's a very big if. Also, most of the research in the field has been done on black and white image, so it not directly usefull. And come to think on it, if we're talking about image stego, what prevents the NSA from controlling the statistical distributions of output of cameras in such a way that stego will be noticed(even after some image transforms) ?
  - IsThisObvious 13y ago
    Maybe the answer is to have a program that adds low scale random noise to as many images as it can before upload to the internet. It would be easy enough to make a one-click batch process, and would help provide cover noise for real communication channels.
zeckalpha 13y ago
Also, this group invented the CAPTCHA.