🚀 go-pugleaf

        Ok - assuming everybody gets a kick at the can -here's mine.

         A search of the literature has unearthed countless references
         to the fact that secret algorithms are out, but no reference
         to an encryption that uses many (known) algorithms, selected
         by the key.

         Assume two reasonably large sets of functions F and G, such
         that Gi reverses Fi for any i. These can be any sort of
         reversible function that operates on a small piece of data
         (eg. a byte). The encryption works as follows:

         1) Use one-way hashes of an arbitrarily long key to produce
         two cryptographically strong keys M and N (A variant of MD5
         that produces longer output?).

         2) Subject the compressed plaintext to a random stub,
         substitution and transposition based on M and/or N to
         complicate a plaintext attack. This produces string D.

         3) Generate E. For each Ei, apply the algorithm from F
         selected by key fragment Mi. Something like

                    Ei = (F(Mi))(N,D,i,E)

        4) Back out step 2. I'm not sure why, but every scheme I've
        seen does this.

        Decryption uses G instead of F and runs i backward.

        This description begs the issues relating to key and block
        lengths, since any scheme can be improved by increasing them.
        The question is, all else being equal, is this an improvement
        over more conventional approaches?

            ,Marc


--
 Marc Thibault        |                              |  Any Warming,
 marc@tanda.isis.org  |  Oxford Mills, Ontario       |  Global or otherwise,
 CIS:71441,2226       |     Canada                   |   appreciated.