[346] in WWW Security List Archive
Re: The Digest
daemon@ATHENA.MIT.EDU (bmanning@isi.edu)
Wed Jan 25 17:21:19 1995
From: bmanning@isi.edu
To: www-security@ns2.rutgers.edu
Date: Wed, 25 Jan 1995 09:35:19 -0800 (PST)
In-Reply-To: <9501251549.AA18552@hopf.math.nwu.edu> from "John Franks" at Jan 25, 95 09:49:52 am
Reply-To: www-security@ns2.rutgers.edu
Errors-To: owner-www-security@ns2.rutgers.edu
In all these activities, you might take a look at the following
internet-draft:
draft-touch-md5-performance-00.txt
Here is the abstract:
ABSTRACT
MD5 is an authentication algorithm, and has been proposed as
one authentication option in IPv6. When enabled, the MD5
algorithm operates over the entire data packet, including
header. We are concerned with how fast MD5 can be implemented
in software and hardware, and whether it supports the network
bandwidth we're installing here at ISI. We have found that MD5
cannot be implemented in existing hardware technology at rates
in excess of 267 Mbps, and cannot be implemented feasibly at
rates in excess of 70 Mbps in special-purpose CMOS hardware.
These rates cannot support ISI's ATOMIC LAN (640 Mbps link, 36
10 Mbps TCP, 75 Mbps UDP, and 300 Mbps native currently). We
believe that if MD5 cannot support existing network bandwidth
using existing technology, that situation will not change in
the future. We propose an alternative to MD5, that is 8- or
16-way block chained, rather than 1-way. We believe that the
resulting algorithm achieves the goals of MD5 over MD4, but
without the serialization penalty that prohibits high speed
implementation. It would require further analysis to ensure
that it provides an adequate level of security.
--bill
