Re: Bank transactions on Internet

["Dave Emery" <die@pig.die.com>]

> 
> At 12:13 AM 4/9/96 -0700, Steve Reid wrote:
> >> Is it really that easy to break 40-bit? Don't you need access to a "fair
> >> amount of cpu power" to brute force crack 40bit? 
> >
> >I remember reading a recent paper at this URL:
> >  http://theory.lcs.mit.edu/~rivest/bsa-final-report.ascii
> >They mentioned a Field Programmable Gate Array (FPGA), specifically a
> >board-mounted AT&T Orca chip available for around $400. They said it could
> >crack a 40-bit key in 5 hours (average). Sounds like anyone with root
> >access on a major internet node could make a significant profit stealing
> >credit card numbers.
> >
> >The FPGA sounds like a very interesting device, with quite a few
> >legitimate uses... Has anyone out there seen one of these? 
> 
> I was hoping a hardware type would answer this question, and give
> references to manufacture's spec sheets, but not having seen such an
> answer, here is a software person's answer.

> 
> 
	As a hardware(and sometimes software) type who has used these sorts
of parts in real designs several things need be said.  

	First, the $400 cost is about what the physical chip and test
board would cost, it does not include the cost of the software packages
required to generate the programming information for the chip and
simulate and verify the design.   While this software can sometimes be
pirated or "borrowed" from an employer or school or even the chip
distributors, charges for a legitimate copy of the software for
programming many kinds of FPGA's can run in the low to mid thousands and
it it is usually dongle protected.

	And the more advanced software packages that take high level
descriptions of the logic in languages such as VHDL and compile them
into the special optimized forms required to get speed out of FPGAs with
highly assymetric routing delays through their interconnect networks are
considerably more expensive and may require RISC workstation hardware
(most of them ran only on Suns or HP in the past) and unix rather than
just a high end PC running Win 95.  Costs of this sort of software
package and workstation run as high as $50K per seat. 

	And it is rather unlikely that one could make a high clock speed
high performance hardware based key cracker work without traditional
high speed logic debugging tools such as a fast logic analyzer (if we
are talking 5-10 ns clock especially) and a 1 ghz or so digital scope.
These kinds of gear, though sometimes available after hours to engineers
working for more liberal companies or schools, cost many thousands
of dollars and are not garden variety items available to any hacker.

	And finally, depending on the technology of the part being
used, there may be a significant cost in the order of at least hundreds
if not thousands of dollars for a specialized programmer capable of
programming ("burning") the FPGA with the interconnect patterns generated
by the software.  These tend to either be specialized to one kind of
part and maybe modestly cheap (hundreds of dollars) or universal and
several thousands of dollars (such as DataIO gear).

	And at least in my experiance (I may be unusually stupid and
careless and clumsy or may not be) even if the parts are a few times
reprogrammable (as CMOS FPGAs often are these days) one can assume
that one will fry, or break the pins off, or reprogram one time too
many the FPGA or FPGA's before one gets the design working.  This
means that it would be realistic to assume several parts would be
consumed by the prototyping effort, they may not be cheap and this
adds up too.

	So whilst someone working with these parts as part of their job
or schooling might well have access to all the required resources on an
informal basis and be able to build a key cracker in evenings or
weekends for little more than the cost of the chip and a PC board to
hold it, it should be realistically noted that the actual cost of
equiping a lab from scratch with the required resources is more on the
order of tens to hundreds of thousands of dollars rather than $400.

	I must hasten to add that high density FPGAs have many many
legitimate uses in prototyping logic and producing products in small
volumes too small to justify the tooling costs of doing mask programmed 
gate arrays (which tend to be significantly faster and easier to design,
but cost $5-100K NRE to set up custom masks for fabrication). The
current generation of them make it possible to build logic systems in
one small chip that a few years ago would have been large  PC boards
full of PALs and other logic.

	Actually designing a workable key cracker for say RC-4 would be
a significant design challenge even with current parts, but probably not
something that someone skilled in the art (and of course reasonably
bright) couldn't handle.  (At first blush I think in the case of RC-4
the pipelined  key scheduling logic required would be the very hard
thing to make efficient).  And the availablity of simulation and timing
analysis tools would make the process of creating such a deamon largely
a software or logic programming exercise that could be mostly carried
out over weeks or months of effort on a workstation or high performance
PC, rather than something that  requires the intensive resources of an
extensive hardware lab for a long period.

	Unfortunately, like so many hacker projects these days, the cost
of reproducing multiple copies of a cracker and the skill level required
is very minimal compared to the real logic programming talent and
architectural insight it would take to fit one into a FPGA or two.
So once one is built, there can be hundreds or thousands of copies
made and put to work in the underground by all sorts of evil people who
wouldn't have a prayer of designing one from scratch,

						Dave Emery
						die@die.com