home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.crypt
- Path: sparky!uunet!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!yale!mintaka.lcs.mit.edu!hal.gnu.ai.mit.edu!knight
- From: knight@hal.gnu.ai.mit.edu (Knight of Tossing and Turning)
- Subject: more mcrypt discussion
- Message-ID: <1992Aug12.185029.13607@mintaka.lcs.mit.edu>
- Summary: more mcrypt discussion (isn't that was title is for?)
- Keywords: mcrypt
- Sender: news@mintaka.lcs.mit.edu
- Organization: It doesn't look like much, but its a system
- Date: Wed, 12 Aug 1992 18:50:29 GMT
- Lines: 98
-
- Mr. Ritter:
-
- Thank you for your reply, I greatly appreciate it (and thank you
- to the other people who have helped out! I take back my previous
- comments.) I've taken this information into consideration and I'm
- going to answer a few of questions and ask a few questions about
- some of your answers.
-
- > If the scheme kept files at *exactly* the same size, I would agree.
- > But if the file size changes (increases) at all, then we cannot
- > (say) simply encipher disk sectors, and what really matters is
- > exactly *how much* the increase is. Perhaps we should consider DES
- > to be equivalently "much more practical" since it also does not
- > expand files (except for blocking and size indication).
-
- The increase is the excess blocking. I encrypt in exactly
- 1024 byte (1K) blocks. I created this method for telephone and disk
- based encryptions -- a "packet" style encryption method.
- The excess is never more than 1023 bytes and can be as little as 0.
- This all depends on (filelength % 1024) and so only that area is
- increased in size. And DES has that problem as well, maybe not to
- the same degree, but due to computer file clusters, this block doesn't
- take up any additional space from the drive.
-
- > >No random numbers will be used anywhere in this encryption.
- >
- > Normally, random numbers are used as part of a cipher system
- > to *solve* serious problems. A problem which comes to mind
- > in this system is whether the internal key-space is provably
- > uniformly distributed among the possible states. If not, then
- > we need to assign individual probabilities to the possible
- > states in order to understand the overall strength, and this
- > can get very complex.
-
- Okay, I can understand this perfectly. I'll write up an explaination
- on how my encryption converts the password line to the 'template'.
-
-
- > * Without specific algorithms to examine, key-space limitations
- > cannot be understood, and this is really the heart of any
- > cipher.
-
- I'm assuming everyone knows what a simple shuffle does. I've already
- included an example of my bit substitution. I'll post the key
- generation method soon.
-
- > * Because of a lack of any statement otherwise, apparently the
- > scheme always enciphers in the same way with the same User Key.
- > If so, this is a serious weakness because it allows anagramming
- > between different messages (and this is the classical method of
- > breaking transposition ciphers). In general, no cipher scheme
- > should "ever" re-encipher the same data in the same way even
- > with the same User Key. (One way around this is to first encipher
- > a "really random" Message Key, and use the plaintext Message Key
- > as the internal key for ciphering the actual data.)
-
- You'll have to explain this to me, because I'm wondering if my
- implementation doesn't completely by-pass this possibility.
- Yes, the 'keys' are the same. They are also disposed and are not
- located anywhere within the file. PASSWORD and PASSWORDPASSWORD
- are not the same, due to the fact password length is included as
- part of the generation of the template (key) but only in manipulation
- and not in deducable mathmatical values (you'll see.)
-
- I've never seen DES encrypt a binary file before, I believe that is
- why you refer to it as "plaintext" cyphers. I could be missing the
- point, but I'm not just encrypting data strings that match ASCII.
- Every attempt made to crack mcrypt will result in a file that will
- be produced as a result of the key. To prove you have 'broken' the
- encryption you MUST create your own way of verifying -- which in
- itself is just a guess at the contents. Even filelength is variable
- because failure to give the right key will cause the 'excess blockage'
- value to be altered.
-
- The basic random influence is "lack of content knowledge" or human.
- If I had a file called "secrets.E" with a filesize of 8192 bytes
- it could be decoded into ascii, zip, arc, arj, tar, shar, pak, zoo,
- Z, z, or even be DES encrypted or encrypted twice with mcrypt.
- This could also be ROT-13 or something just as cheezy.
-
- Even were there as few as a million permutations, how do you
- determine if its right? This is a bad excuse, as I was trying
- to gain permutations to the point that technology won't catch
- up with this system any time during my lifetime. I'm still
- confident there are more than DES.
-
- Yes, I was attempting to create an encryption that would become
- I/O bound as the limit to its speed. Most other systems become
- CPU bound (DES, great example.) I/O is significantly slower than
- CPU and that causes problems repeatitively trying to guess it.
-
- > ---
- > Terry Ritter ritter@cactus.org
-
- -- Eric Knight
- knight@gnu.ai.mit.edu
-
-
-