Apology and clarification

First of all, I believe that I owe the cypherpunk community an apology
for an error in judgement on my part.  The message that I sent out
yesterday regarding our demonstrations of a newly-discovered security
threat was the exact same text that I  sent to a  far less technical
audience.  As such, I understand that many people on this list found the
tone of my message to be insulting and offensive.  I apologize, and I
certainly didn't mean to insult anyone's intelligence.  

Having said that, please cut me a break.  If you read my message as
saying "FV has just invented keystroke sniffing" you've completely
missed the real attack here.  If you really think I'd throw away my
reputation on a bogus claim like that, you're insulting *my*
intelligence.   My (charitable?) take on it is that a lot of people were
so put off by the tone of my mass-market message that they leapt to the
quick but erroneous conclusion that there was no underlying content. 
There is.

The threat is NOT from keystroke sniffing per se, and we're certainly
not claiming to have invented keystroke sniffing.  However, we do have
to *explain* keystroke sniffing in the public announcement, because it
is a *part* of our attack, and most of the public does NOT already know
that it's possible.

What we at FV have done is to demonstrate how easy it is to develop an
FULLY AUTOMATED attack that undermines the security of all
software-based credit card commerce schemes.  It is the automated aspect
that separates it from all of the "dumpster-diving" attacks on credit
card numbers which have previously been widely discussed, because it
provides a path to large-scale fraud that has never been publicly
discussed before, to my knowledge.  The key "invention" in our approach
is to integrate several techniques that are already well-known (in this
community) into an automated attack that we consider to be devastating
to commerce systems based on software-encrypted credit cards.

Our approach combines the following four known problems into a fatal attack:

	1) Consumer machines are insecure and easily compromised.
	2) Keyboard sniffers are easy to write.
	3)  Credit card numbers are self-identifying (they have check digits) 
	    and can easily be extracted from a huge stream of input data.
	4)  Once intercepted, small amounts of information (e.g. a cc #)
	     may be distributed completely tracelessly over the Internet.

When you put all four of these together, you have an attack that IS new,
in the sense that nobody we know of has ever mentioned it before, and
which could in fact be used by a single criminal, with only a few weeks
of programming, to tracelessly steal MILLIONS of credit cards, if
software-encrypted credit-card schemes ever caught on.

This is a very real threat.  If you think we're just re-hashing keyboard
sniffers, you haven't yet understood what we're demonstrating.  The real
threat is the traceless theft of millions of credit card numbers by a
single easily mounted automated attack.

So here's the factual claim, to be proven or disproven:  One good
programmer, in less than a month, can write a program that will spread
itself around the net, collect an unlimited number of credit card
numbers, and get them back to the program's author by non-traceable
mechanisms.  Does anyone on this list doubt that this is true?  If so,
I'd like to know the flaw in my thinking, -- I am *not* too proud to
withdraw any claims that aren't true.  If not, I think it's worth noting
that this fact was previously completely unknown to the bankers and
businessmen who are putting large sums of money at risk on the net.  The
only way to get the message to those communities is with a very visible
public announcement of the kind you saw yesterday.  -- Nathaniel
--------
Nathaniel Borenstein <nsb@fv.com>
Chief Scientist, First Virtual Holdings
FAQ & PGP key: nsb+faq@nsb.fv.com

Excerpts from mail.cypherpunks: 30-Jan-96 Re: Apology and clarification
Jamie Zawinski@netscape. (4170*)

> Nathaniel Borenstein wrote:
> > 
> > What we at FV have done is to demonstrate how easy it is to develop an
> > FULLY AUTOMATED attack that undermines the security of all
> > software-based credit card commerce schemes.

> You have done no such thing.  You have written *one component* of that
> attack, and the easiest part of it at that.

> Combine it with a virus, or self-replicating worm, and demonstrate that
> it is immune to all known virus checkers, and *then* you will have
> spoken the truth when you say you have "demonstrated" anything.

This is a particularly fascinating reaction, Jamie.  As I see it, we
have implemented every part of the attack that we can implement without
doing anything that is either unethical or illegal.  Is it your position
that no systematic flaw in your security is real until someone has
actually broken it?  

Actually, that position would in fact be quite consistent with your
company's earlier implicit assertion that 40-bit encryption was
sufficient (for international consumers) until somebody actually broke
it, even though everyone who understood cryptography already knew
otherwise.

> You may think this is nitpicking, but the fact is, you're assuming that
> the implicit cooperation of some vast number of users in running your
> program is easy to obtain.  I disagree with this assumption.  If this
> assumption were true, then viruses would be a much bigger problem than
> the mere annoyance that they are today.

Nearly everyone with a computer has either been infected with a virus or
knows somebody who has.

There has never been a serious financial incentive for virus writers in
the past, so they haven't ever been, for example, bankrolled by
organized crime.  They've been written by sociopathic hobbyists in the
past.  Your commerce mechanism gives them an incentive to turn pro.

The average sophistication of Internet users is dropping every day, as
the net continues to explode, and the ease of spreading malicious
software is going up accordingly.

Having said all that, I do agree with you 100% that the hardest part of
the devastating, automated attack that we have outlined is in fact the
infection vector.  You are absolutely right about that.  What we have
shown is that the HARDEST part of stealing an unbounded number of credit
cards transmitted using your company's preferred commerce mechanism is,
in fact, the deployment of a virus or Trojan Horse.  Unfortunately, as
most personal computer users have long since realized, that just isn't
that uncommon or hard to do.

> *Computers* provide a path to large-scale fraud.  So does the printing
> press.  So does the telephone, and the postal system.  So what.  You
> still haven't proven that it's easy.

I suspect that the world's financial institutions will, by and large, be
grateful that First Virtual doesn't share your belief that one has to
wait for a criminal to break a system to be convinced that it is
insecure. 

Show me an automated way to break the postal system in a large-scale way
without getting caught, and then I'll be worried about it, too.

> With as much work as you've put into this, someone could write a
> Microsoft Word document which when opened, would start dumping the
> contents of your hard disk into the mail.  

Ooh, good point.  We could probably use MS Word macros as the infection
vector for our program.  I like that idea.  I'll add it to our list of
potential ways this program could spread itself.

However, the entire contents of your hard disk aren't of direct economic
value.  They're also hard to digest, and they're big enough to be likely
to be noticed in transit (e.g. they can easily fill up mail spools if
you mail 'em out).  I'd much rather sift through your hard disk looking
for credit card numbers, and then spirit them quietly off your machine. 
But I'd also install a keystroke sniffer if I suspected the user might
be using your preferred mechanism to send out his credit card number.

> It's not a matter of possibility.  It's a matter of probability, and
> risk management.  It's unlikely enough that I'm not afraid of using my
> credit card on the net.  Tell me my credit card number, and I'll change
> my mind.

Hey, you're a smart guy.  That probably means your machine is relatively
hard to infect.  A criminal would skip you and instead target the
millions of consumers who were more easily infected.  I didn't describe
a scheme that could target one individual's credit card.  I described a
scheme that could steal millions of them indiscriminately.

> All a banker needs to know is the amount of risk associated with the
> thing in which they are investing; they don't need to know how keyboard
sniffers work.

The "trust us, we're experts" approach to security is only as good as
the experts you trust, as you've just amply demonstrated.  For my part,
I'm happy to let the bankers hire independent experts study the attack
we've outlined and reach their own conclusions.  -- Nathaniel
--------
Nathaniel Borenstein <nsb@fv.com>
Chief Scientist, First Virtual Holdings
FAQ & PGP key: nsb+faq@nsb.fv.com

Nathaniel Borenstein wrote:
> 
> What we at FV have done is to demonstrate how easy it is to develop an
> FULLY AUTOMATED attack that undermines the security of all
> software-based credit card commerce schemes.

You have done no such thing.  You have written *one component* of that
attack, and the easiest part of it at that.

Combine it with a virus, or self-replicating worm, and demonstrate that
it is immune to all known virus checkers, and *then* you will have
spoken the truth when you say you have "demonstrated" anything.

Heck, combine it with a screensaver as a trojan horse *and* collect a
few hundred credit card numbers and *then* you will have demonstrated
something.

You've demonstrated nothing but your ability to write press releases,
and print out some messages when fully-cooperating users submit to your
"test."

You may think this is nitpicking, but the fact is, you're assuming that
the implicit cooperation of some vast number of users in running your
program is easy to obtain.  I disagree with this assumption.  If this
assumption were true, then viruses would be a much bigger problem than
the mere annoyance that they are today.

> It is the automated aspect that separates it from all of the
> "dumpster-diving" attacks on credit card numbers which have previously
> been widely discussed, because it provides a path to large-scale fraud
> that has never been publicly discussed before, to my knowledge.  The
> key "invention" in our approach is to integrate several techniques
> that are already well-known (in this community) into an automated
> attack that we consider to be devastating to commerce systems based on
> software-encrypted credit cards.

This is the same kind of vaccuous reasoning that leads to things like
the "concryption" patent.  You have invented nothing.  You've combined
the painfully obvious and written a fearmongering rant about it.

*Computers* provide a path to large-scale fraud.  So does the printing
press.  So does the telephone, and the postal system.  So what.  You
still haven't proven that it's easy.

> This is a very real threat.  If you think we're just re-hashing keyboard
> sniffers, you haven't yet understood what we're demonstrating.  The real
> threat is the traceless theft of millions of credit card numbers by a
> single easily mounted automated attack.

With as much work as you've put into this, someone could write a
Microsoft Word document which when opened, would start dumping the
contents of your hard disk into the mail.  

The knee-jerk moral to *that* is to never store non-public information
on a computer that has a network connected to it.

However, reasonable people assess that risk, and decide to do it anyway,
because the benefits outweigh the risk.

> So here's the factual claim, to be proven or disproven:  One good
> programmer, in less than a month, can write a program

Come now, right off the bat you know that no assertion taking that form
can be *dis*proven.

> that will spread itself around the net, collect an unlimited number
> of credit card numbers, and get them back to the program's author by
> non-traceable mechanisms.  Does anyone on this list doubt that this 
> is true?

It's not a matter of possibility.  It's a matter of probability, and
risk management.  It's unlikely enough that I'm not afraid of using my
credit card on the net.  Tell me my credit card number, and I'll change
my mind.

> If not, I think it's worth noting that this fact was previously
> completely unknown to the bankers and businessmen who are putting
> large sums of money at risk on the net.  The only way to get the
> message to those communities is with a very visible public
> announcement of the kind you saw yesterday.

All a banker needs to know is the amount of risk associated with the
thing in which they are investing; they don't need to know how keyboard
sniffers work.  I don't believe you've demonstrated anything that
changes the risk model that they have presumably already gotten from
their flock of experts who they no doubt employed before investing in
the net (experts who also no doubt know all about how viruses work,
thank you very much.)

	== Jamie

Excerpts from mail.cypherpunks: 1-Feb-96 Re: C'mon, How Hard is it t..
Jamie Zawinski@netscape. (2014*)

> > Is it your position that no systematic flaw in your security is real
> > until someone has actually broken it?

> Of course not.  You don't have to actually break it to show that it's
> possible.

> Of course, you *do* have to show the likelyhood of success and effort
> required to pull it off as well before it's interesting at all, whether
> it's theoretically possible or not.

OK, let's try this again:  Is it your position that the hardest part of
the attack we've outlined is the large-scale infection of consumer's
machines with untrusted code, using a virus, Trojan Horse, or some other
method?  And that this attack is not serious because doing that is
prohibitively difficult?  If so, I agree with the first claim but not
the second.  But I'm really trying to get clear about your position
here.  -- Nathaniel
--------
Nathaniel Borenstein <nsb@fv.com>
Chief Scientist, First Virtual Holdings
FAQ & PGP key: nsb+faq@nsb.fv.com

Jeff Weinstein wrote:
> 
> I think that you are misinterpreting the intent of Jamie's posting,
> but I will let him defend himself.

Well I'm not particularly interested in arguing about this further
(and I suspect this is true of most people reading this too :-))
but my point was: Nathaniel and crew have implemented the easy part 
(a tiny fraction) of a program which would successfully capture some
large number of credit card numbers.  Nathaniel thinks that what I'm
characterizing as a tiny fraction of the work (the keyboard sniffer and
pattern recogniser) is *most* of the work, and "demonstrates" the
attack.  I said that they have demonstrated nothing without some proof
that combining this with an infection vector would yield the desired
result, because I don't think that infecting some vast number of
credit-card-using computers is any small task; whereas, Nathaniel says
(or at least strongly implies) that it's trivial (or so close to trivial
that it can be taken as a given.)

Nathaniel said:
> As I see it, we have implemented every part of the attack that we can
> implement without doing anything that is either unethical or illegal. 

It's far from clear that you need to do something unethical or illegal
to prove that coupling it with an infection vector would be effective.

For example, you would no doubt agree that evesdropping on some
unsuspecting user's transaction on an exportably-crippled SSL connection
would be immoral.  But it wasn't necessary to do anything immoral to
demonstrate conclusively that such an attack was possible.  It just
required a little creativity, and a lack of handwaving.

> Is it your position that no systematic flaw in your security is real
> until someone has actually broken it?

Of course not.  You don't have to actually break it to show that it's
possible.

Of course, you *do* have to show the likelyhood of success and effort
required to pull it off as well before it's interesting at all, whether
it's theoretically possible or not.

	== Jamie

Nathaniel Borenstein wrote:
> 
> Excerpts from mail.cypherpunks: 30-Jan-96 Re: Apology and clarification
> Jamie Zawinski@netscape. (4170*)
> 
> > Nathaniel Borenstein wrote:
> > >
> > > What we at FV have done is to demonstrate how easy it is to develop an
> > > FULLY AUTOMATED attack that undermines the security of all
> > > software-based credit card commerce schemes.
> 
> > You have done no such thing.  You have written *one component* of that
> > attack, and the easiest part of it at that.
> 
> > Combine it with a virus, or self-replicating worm, and demonstrate that
> > it is immune to all known virus checkers, and *then* you will have
> > spoken the truth when you say you have "demonstrated" anything.
> 
> This is a particularly fascinating reaction, Jamie.  As I see it, we
> have implemented every part of the attack that we can implement without
> doing anything that is either unethical or illegal.  Is it your position
> that no systematic flaw in your security is real until someone has
> actually broken it?
> 
> Actually, that position would in fact be quite consistent with your
> company's earlier implicit assertion that 40-bit encryption was
> sufficient (for international consumers) until somebody actually broke
> it, even though everyone who understood cryptography already knew
> otherwise.

  Actually that position would in fact be quite inconsistent with our
more recent actions.  For example we have implemented blinding code to
protect against Paul Kocher's timing attack, even though it has
not been demonstrated against any real world system.  I think that you
are misinterpreting the intent of Jamie's posting, but I will let
him defend himself.  I just wanted to say that the company takes
security problems very seriously, even if there has not been an
active exploit.

	--Jeff

-- 
Jeff Weinstein - Electronic Munitions Specialist
Netscape Communication Corporation
jsw@netscape.com - http://home.netscape.com/people/jsw
Any opinions expressed above are mine.