Re: How much must be revealed


On Mon, 26 Feb 2007 HilltopLab@xxxxxxxxxxxxx wrote:
On Feb 26, 9:35 am, Volker Hetzer <firstname.lastn...@xxxxxxxx> wrote:
Hilltop...@xxxxxxxxxxxxx schrieb:> If I have an encryption program, is it possible to describe the
algorithm sufficiently to give people confidence in its security
without revealing the source code? Or must it be given away before any
worth is appreciated?

These are two different issues:
You can convince peope to trust your /algorithm/ by publishing it. Or by
choosing a well known one and saying "I use AES/CTR with that kind of
Key/Counter management".

You can convince people that you are not a liar by providing source code.
In cases without random numbers (i.e. if your program is fully deterministic)
people can perhaps trust you if they can reproduce ciphertext by means of
pencil and paper and it matches your output and you take reasonable pains
to ensure integrity of thge download (i.e. by providing pgp keys, checksums
and the like).

I completely agree that you cannot trust a program that depends on the
secrecy of the algorithm. Not only because the algorithm cannot be
independently tested, but also because no one can ensure the secret
will not get leaked. But that's why I asked the question. No matter
how detailed an explanation is given of an algorithm, the devil is in
the details. And, just because I claim the program does something
doesn't mean it actually does. So, if the only recourse is to publish
the source code, there is little incentive to develop a new algorithm,
as no one can hope to market it. Except maybe as a hobby.

Did you read my response to your original post? In part:

They just have to trust you --- maybe because you're a well-known good guy, or maybe because you offer a comprehensive warranty (hah!), or maybe because they're just stupid or lazy (likely).

There are plenty of stupid and/or lazy people out there to keep even
Microsoft-built crypto software in the black. And there are a fair number
of well-known good guys who provide crypto software, although admittedly
they do go the open-source route.
However, nobody honest makes money on copies of a single program
anyway. You make money on the support contracts, or on custom
modifications, or things of that nature. And that business model
applies just as well to crypto as to other fields --- possibly better,
because crypto is so hard to get right. (Or possibly worse, because
so many crypto customers are idiots *cough*MPAA*cough*, and those who
aren't idiots would tend to have in-house crypto programmers and not
contract out so much, I'd expect. Hmm.)

If your business proposition is essentially "Hey, look, I made an algorithm!", then yeah, you're not going to make any money.

HTH,
-Arthur
.

Relevant Pages

  • Re: How much must be revealed
    ... You can convince peope to trust your /algorithm/ by publishing it. ... You can convince people that you are not a liar by providing source code. ... Not only because the algorithm cannot be ...
    (sci.crypt)
  • Re: Encryption key changing the encryption logic.
    ... >>decides among various elementary crypto operations based on what it sees ... >>adversary knows which key bits select which algorithm and, in practice, N ... > or two rounds of a different block cipher. ... this design is covered under the 1-of-N discussion and your design is ...
    (sci.crypt)
  • Re: Is it possible to encrypt without a key ?
    ... Start by learning about crypto, you can get the Handbook of Applied ... Encryption, with a fairly good algorithm, is usually a fast process ... >extractor easily ... >generic extractor by desassembling any extractor of any master. ...
    (sci.crypt)
  • Re: Possible new crypto system
    ... Standard cryptographic advice is never to touch any secret algorithm ... if I had a way to encrypt a "text" (some binary data ... Those are the basic requirements for any crypto system. ... Any good modern crypto system is proof against brute force. ...
    (sci.crypt)
  • Re: Blowfish Sign Extension implementation risk
    ... > Tom St Denis wrote: ... >> the algorithm correctly in the first place is the better course of action. ... mainstream, was published with C source code (again, see Dr. Dobbs, ... Applied Cryptography, etc.). ...
    (sci.crypt)