Other Directory Sites: SeekWonder | Directory Owners Forum
The following archive was created by hippie-mail 7.98617-22 on Sat Apr 10 1999 - 01:18:49
David R. Conrad (drc@adni.net)
Mon, 1 Mar 1999 07:04:23 -0500 (EST)
On Thu, 25 Feb 1999, Vin McLellan wrote:
> Thomas Roessler <roessler@guug.de> queried the List:
>
> >Are there any current estimates on the cost and time it takes to
> >factor a 512 bit RSA modulus?
>
> Time and money (CPUs, memory, and talent) obviously play off
> against one another in such a task.
>
> Jim Gillogly <jim@acm.org> responded:
>
> >Paul Leyland ... expects such a project to be completed
> >within a year or two.
>
...
> Projections now give us a fairly accurate estimate of the amount to
> processing time and memory required to tackle such a task like RSA-512
> using GNFS.
>
> With similar pre-processing to select the polynumerals, and with
> access to a similar array of computational resources, RSA estimates that
> factoring a 512-bit (155 digit) RSA modulus should take about 7.2 times as
> much processing time as the RSA-140 project. That's 7.2 months on the same
> equipment: 64 CPU-years. (Sorry. I can't independently calculate the
> MIPS-year equivalent; admittedly a more useful stat.)
>
> Holding the matrix for a NFS attack on a 512-bit RSA-155 integer is
> also expected to require about 2.7 times as much memory as the 810 MB in
> C916 main memory that was used when they factored RSA-140. That's a
> whopping 2,187 MB.
http://www.altavista.com/av/content/freshindex.htm mentions "state-of-the-
art two million dollar Alpha 8400 computers with 6 to 8 Gb of RAM and up
to 400 Gb of disk space each". I wonder if the folks at AltaVista could
be persuaded to participate in an effort to break such a number.
(I went looking for the above stat because I remember being impressed some
time ago by the ludicrous amounts of memory used by some of the AltaVista
machines; I don't know if that's actually 6-8 gig of memory per machine,
or total -- "each" clearly refers to the disk space, but whether it also
applies to the memory is ambiguous -- damn natural languages!)
> (By contrast -- again using the stats from the RSA-140 Challenge as
> a baseline -- factoring a 1024-bit number is roughly estimated to require
> about 40 million times more processing, and 6,300 times as much memory.)
That, at least, should be safe for a while yet.
David R. Conrad <drc@adni.net> PGP keys (0x1993E1AE and 0xA0B83D31):
DSS Fingerprint20 = 9942 E27C 3966 9FB8 5058 73A4 83CE 62EF 1993 E1AE
RSA Fingerprint16 = 1D F2 F3 90 DA CA 35 5D 91 E4 09 45 95 C8 20 F1
I will show you fear in a handful of cruft.
The following archive was created by hippie-mail 7.98617-22 on Sat Apr 10 1999 - 01:18:49