From: edgarswank@juno.com (Edgar W Swank)
Date: Mon, 23 Dec 1996 04:38:20 -0800 (PST)
To: Cypherpunks          <cypherpunks@toad.com>
Subject: Re: Blowfish Performance
Message-ID: <19961223.043602.2974.1.edgarswank@juno.com>
MIME-Version: 1.0
Content-Type: text/plain


-----BEGIN PGP SIGNED MESSAGE-----

David Stoler recently commented:

>I ftp'd the Blowfish implementation from ftp.ox.ac.uk and ran some
>benchmarks.
>
>Encryption and decryption are faster than the fastest DES implementation
>I've found, but the key expansion (subkey generation) is slow.
>
>On a 100 Mhz PowerPC I get the following results:
>
>Key expansion:
>350 keys per second = 3.33 milliseconds per key
>
>Encryption/Decryption:
>190,000 8 byte blocks per second = 5.26 microseconds per block
>
>This is 40% faster (encryption/decryption) than the best DES I have.
>
>Questions:
>...
>3.  Is there a way to improve key expansion performance for applications
>that change keys frequently?  (Other than caching expanded keys...)

Actually this (slow key expansion) is a feature. It limits the
speed of a brute-force attack, since each new key to be tried
must be expanded.

What applications need to change keys more than once a second?

I'd also be interested in comparisons of Blowfish to IDEA.

Edgar W. Swank   <EdgarSwank@Juno.com>
                 (preferred)
Edgar W. Swank   <Edgar_W._Swank@ilanet.org>
Edgar W. Swank   <EdgarSwank@Freemark.com>
Home Page: http://members.tripod.com/~EdgarS/index.html

-----BEGIN PGP SIGNATURE-----
Version: 2.6.2

iQCVAwUBMr4fSt4nNf3ah8DHAQHycwP7BWTgyHVVZrtEhlMQKmSFFQGcMKqxaa5V
y7D0OSVGTuddkwFDsVR7Qt3YGV0q3sMWbXFbicn2R/25rBbFNzA8d2rdgIxJVEzZ
0uKb1xdWkVR7GLIlQDciQW/zdQxdXjkiQ9sVbUJtPwYzUdZt6us3Izac1FZ4fKAX
mvlqa4yWG58=
=Tmc6
-----END PGP SIGNATURE-----