Cryptography Stack Exchange is a question and answer site for software developers, mathematicians and others interested in cryptography. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm developing a new image steganalysis technique, and I'd like to test how effective it is in practice. How could I do this? Are there such things as standard "test vectors" for steganalysis?

share|improve this question

closed as off-topic by e-sushi, Gilles, DrLecter, woliveirajr, tylo Dec 17 '13 at 12:46

This question appears to be off-topic. The users who voted to close gave this specific reason:

If this question can be reworded to fit the rules in the help center, please edit the question.

This isn't a cryptography question. Perhaps – John Deters Dec 13 '13 at 13:50
Requests for reference recommendations are off-topic here. For details, see: Do we want “literature recommendations” and similar “list/subjective questions”? – e-sushi Dec 13 '13 at 20:40
I'm not sure if this question really qualifies as a "reference recommendation" in the sense discussed at the meta post @e-sushi linked to. Even if it does, though, rephrasing it e.g. as "How can I test the effectiveness of my new image steganalysis technique on realistic data?" should make it on topic here. – Ilmari Karonen Dec 14 '13 at 14:19
@Layla: I just tried to do that myself, but feel free to change anything you want (especially if you feel I introduced any inaccuracies). It's your question, after all. – Ilmari Karonen Dec 14 '13 at 14:28
Testing a new CSPRNG's output with existing randomness tests is next to pointless in order to show the goodness of the CSPRNG (tests can only detect some very bad CSPRNG, and some implementation errors). Similarly, testing a steganalysis technique with existing test data (not designed with knowledge of the steganalysis technique) will not be a good demonstration of the virtues of the new techniques. – fgrieu Dec 14 '13 at 14:59

Are there such things as standard "test vectors" for steganalysis?

Not that I am aware of.

How could I do this?

Your best bet would be to look for steganographic systems that are out there (either software or a paper describing it). Implement those systems and see how your steganalysis method works. Then you can implement other steganalysis methods from the literature and compare how well yours does vs those.

share|improve this answer

Not the answer you're looking for? Browse other questions tagged or ask your own question.