🚀 go-pugleaf

unruh@physics.ubc.ca (William Unruh) writes:
: who@gandalf.UMCS.Maine.EDU (William H. Owen) writes:
: 
: 
: >	There is a nice version of DES in pascal included in "Computer Networks"
: >by Andrew S. Tanenbaum. Anyone have a version of DES with a 128 bit key size? Or
: >alternative choices for S-boxes? Or just maybe the NSA doesn't want you to crypt
: >something that the gov't can't uncrypt.
: Uh, read your first line. You have the source- change the Sboxes to your
: heart's content. 

NO!  Don't change the S-boxes or the ordering of the S-boxes because
it greatly weakens DES.

: Double, triple, qudruple encrypt with different keys,
: and different permutations. If NSA had really wanted to make sure you
: couldn't use anything but their weak(???) DES, they would never have
: agreed to release the details of DES for you to change them.

Over the years, DES has proved to be very resistant to attempts to
crack it.  Many "improvements" have been proposed which have later
been broken.  Unless you are an expert cryptologist you shouldn't
change the algorithm at all.  

By all means use multiple encryption, though.  Double encryption
probably isn't worth bothering with because of "meet in the middle"
attacks.  Two key triple encryption is a good bet.  This is done by:

    plaintext -> enc(a) -> dec(b) -> enc(a) -> ciphertext

where enc(a) is enciphering with a, dec(b) is deciphering with b.  The
reason for this arrangement is that if the keys a and b are the same,
this process is equivalent to single key encryption.  This is
convenient for compatibility with other (single key) programs.

There is mounting evidence that IBM, in conjunction with the NSA,
really did design a secure algorithm.  I realise that this won't
satisfy conspiracy theorists, but then, by definition, they won't ever
be satisfied, however much evidence there is.

Andrew.