Non-Disclosure Agreement

Non-Disclosure Agreement

Don't Disclosure Agreement

Just a friendly reminder - do not post videos, images, stream, or any other content from our alpha. It is against the NDA that you agree to prior to joining us in testing. If you have any questions in regards to the NDA - please post them here.

>>3
The problem these days is that everything is watermarked and signed. So when you do leak, when know who did.

But the advent of F.L.O.S.S. changed most of that, and SN.

Still, /prog/iders shoudl be able to de-sign and reprint

>>4

It is easy to detect and remove watermarks:
1. get another person with NDA to send you CRCs of his files
2. see what files have different CRCs
3. strip them

>>5
There could be multiple watermark modifications; you and your buddy can't detect watermark modifications that are common to both of you. An analyst can then easily figure out who were the traitors. Enjoy your daily prison rape, federal criminal.

>>5

Still, /prog/iders shoudl be able to de-sign and reprint

I am not saying it's not impossible, but generally binaries are riddle with multiple watermarks, even some in plain sight, like that one pixel or double byte with your GUID. It's best not to on corporate scenarios. It's why in indies I make the big bucks

>>6,7
Just normalize the data. Convert pdfs to text files. Run audio through a filter that will mangle anything not human detectable. Run similar filters on images. Blur pixels. Add noise. Decompile binaries, eliminate dead code, and recompile. When recompiled, make the code execute itself by decrypting itself multiple times in 256 different ways.

Run audio through a filter that will mangle anything not human detectable.

Heh, and get bad audio in return.

on images. Blur pixels. Add noise.

So, subpar work?

eliminate dead code, and recompile.

I make sure important variables have your signature GUID, usually on RGNs

When recompiled, make the code execute itself by decrypting itself multiple times in 256 different ways.

Explain. Not all executables are compressed.

>>9
Not all ani are compressed either. That doesn't stop me from compressing them.

>>6

Hey! You just gave me idea for AI.

>>10
LZO my ANUS

>>7
Consider a software distributed to N different users with each distribution having N*N bits watermark, where each user's distribution shares N bits with other N-1 users. In such situation malicious users A and B can't detect watermark by comparing their distributions, because such comparison won't show the bits they share and won't expose other bits, allowing identifying the leaker uniquely. You can easily show that if n users can't detect watermark, then n+1 users also cannot.

>>8

Run audio through a filter that will mangle anything not human detectable.

Won't help. See http://en.wikipedia.org/wiki/Canary_trap

>>14
All those techniques are just as easy to circumvent as they are to implement. If they use synonym shuffling, then perform your own synonym shuffling on the document. You just need to anticipate every way they will try to encode the watermark and you'll be able to noise it out.

>>13
You don't need to recover all the bits of the watermark. You just need to modify all of them, possibly modifying other bits as well.

>>15

You don't need to recover all the bits of the watermark. You just need to modify all of them, possibly modifying other bits as well.

You can't know all of them.

>>16
You can know for certain if you compromise the canary trap department. But even if you don't you can run the original through enough transformations where recovery of the original watermark would take more resources than would be worth the effort. If you are lucky, all traces of the watermark are obliterated. If some remain, then you would hope that the distortion modified the watermark to the point where it can't be distinguished from it's other versions. Depending on the transformation, this can be high probability. The transformation would need to be suited for the watermark. In the worst case, the watermark is the content of the document itself, and the different versions given to the privileged members are all completely different. In this case the only safe transformation is one that reduces the text to pure noise and destroys all the content. Ergo, you're wrong. If I xor every bit of the document with a random bit array you'll never know what it was originally and the leaker is safe. Unfortunately the content of the document is also destroyed.

/prog/ CHALLENGE #665.9
Write a program that is capable of rewriting a document to a semantically equivalent document with a minimal amount of alterations that alter meaning in a way that cannot be inferred by a human reader.

BONUS: implement as many AI\machine learning cargo cult techniques as possible.

>>18

that cannot be inferred by a human reader.

Impossible.

>>19
Interprete that to mean `by a non-retarded\non-lawyer human read that can tell what was meant by a misused word from the context'.

>>18

BONUS: implement as many AI\machine learning cargo cult techniques as possible.

Ridiculing machine learning will get you nowhere, because people are making serious buck of AI since the inception of the term "AI".

_{(stopping the dubsfaggot from dubsbumping)}