# [rc5] Question for beginner

Eric Gindrup gindrup at okway.okstate.edu
Fri Oct 24 12:53:09 EDT 1997

```        I see that your question of how much larger the keyspace is has
hasn't been deeply responded to.
The just finished contest has the full name RC5-32/12/7 and the new
RC5-64 has the full name RC5-32/12/8.  The number of key bits is
determined by multiplying the last number in the set of three by
eight.  So, 7*8 = 56 for the previous effort and 8*8 = 64 for the
current effort.  The other two numbers are (first) "word size" and
(second) "rounds".
RC5 decryption occurs in a couple of stages.
1.  Expand the key into a bunch of pieces, 2 for each round plus 2
more.
2.  Do some scrambling of the ciphertext for each round using the
pieces from step 1 in pairs.  (this is why you need twice as many
pieces)  Then do a little (irrelevant) scrambling one last time (thus
using the "plus 2" last pieces).

Step 1 is pretty straightforward and should only require recoding
when the number of bits in the key gets larger than whatever processor
register is being used to hold those values.  So the jump from 56 to
64 bit keys won't change this much.  Beyond 64 bit keys will probably
require a signficant speed hit as pieces of the key will have the be
swapped into and out of the processor...
Step 2 is similar for different key lengths, but some
implementations represent the 32bit words as a collection of little
pieces and perform operations on each piece separately.  So, if you
increase the number of pieces, you have to perform more (of the same)
operations on the new pieces.

Strictly, the changes are relatively straightforward on a clean
implementation.  Highly optimized implementations may be very
difficult to modify in this way.  This is especially true since most
of the changes are of the "do what you've been doing just one more
time" sort.  That these changes aren't *very* drastic is evinced by
the rapidity with which the 56bit clients were refitted to handle
64bit keys.

Abstractly, the algorithm has only changed in one parameter: the
key length.  And that parameter doesn't affect the abstract algorithm,
only its implementatino on real hardware.

I'm sort of curious if the new clients post about 88% performance
of the old clients (in keys/sec) due the the 7/8 utility multiplier...
-- Eric Gindrup ! gindrup at Okway.okstate.edu

Subject: [rc5] Question for beginner
Author:  <rc5 at llamas.net > at SMTP
Date:    10/24/97 1:41 AM

Ok, here is my question. What is changing from moving from the 56 bit
key to the
64?
I know the keyspace expands a lot -how much, really?
But then the algorithm remains the same isn't it? Why we need new
clients?

Thanks a lot, Harris
----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the
body.

----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.

```