The Risks Digest

The RISKS Digest

Forum on Risks to the Public in Computers and Related Systems

ACM Committee on Computers and Public Policy, Peter G. Neumann, moderator

Volume 23 Issue 16

Tuesday 3 February 2004


Security holes at DMVs nationwide lead to ID theft and safety concerns
Monty Solomon
Defeating phishing scams
Andrew Rose
A nasty Phishing attempt
Avishai Wool
Another wireless risk
Chris Meadows
Hotel reservation system easily confused
Richard S. Russell
Browsers, online forms, rendering and opt-in marketing
Alistair McDonald
Drunk unlocks police car with own key
Re: Happy 2**30'th birthday, time_t!
Steve Summit
Re: Suing the customers
Paul Robinson
Re: Lie-detector glasses, 90% accurate?
Ron Bean
Peter B. Ladkin
REVIEW: "Biometrics", Woodward/Orlans/Higgins
Rob Slade
Info on RISKS (comp.risks)

Security holes at DMVs nationwide lead to ID theft and safety concerns

<Monty Solomon <>>
Mon, 2 Feb 2004 09:50:52 -0500

Security Holes at DMVs Nationwide Lead to ID Theft and Safety Concerns

CDT has issued a report entitled "Unlicensed Fraud" documenting rampant
internal fraud and lax security at state motor vehicle administration
offices across the country placing the reliability of all driver's license
at risk. While heavy public attention has been placed on new national
standards and new technologies for driver's licenses, studying local news
reports from throughout 2003 CDT finds that basic management processes to
stop bribery and theft are lacking. In the report, CDT offers policy
recommendations to address this dire issue.  2 Feb 2004

Unlicensed Fraud: How bribery and lax security at state motor vehicle
offices nationwide lead to identity theft and illegal driver's licenses:

Defeating phishing scams

<Andrew Rose <>>
Mon, 19 Jan 2004 10:11:35 -0000

PGN commented on another 2 phishing scams highlighted in RISKS-23.13,
> [This is increasingly becoming a problem!  We desperately need
> some greater authentication and accountability.  PGN]

Work on a technical (part-)solution named SPF ("senders permitted from") is
underway at  This simple technique has domains
publish so-called SPF records in the DNS.  The SFP records detail those
machines that may validly send email for the domain in question.  This
allows receiving MTAs to reject or flag email that claims to come from e.g. but isn't sent by a machine that is authorised to send on behalf
of paypal (e.g. a phisher).

The technical work on SPF is now complete and adoption has started.  Several
thousand domains have published SPF records including some very large
domains such as  Plugins exist for most of the popular MTAs - the
only notable exception being MS Exchange.

For a more detailed overview see  Those who are still
interested should then read and join the mailing list.

A nasty Phishing attempt

<Avishai Wool <>>
Sat, 24 Jan 2004 09:12:42 -0800 (PST)

Perhaps this is ho-hum for RISKS readers, but I thought I'd pass this along

I got a nasty spam today which I excerpt below.
It purports to be from the FDIC, and asks the reader
to go to the FDIC web site and "verify" bank account details.
It uses the
trick, where the "reasonable" site name is treated as a username.
The criminal site probably attempts to harvest these details (I tried it
but the site was unresponsive).

This is a clever piece of social engineering, which is especially
effective against non-US residents that have (or had) a US bank account.


> Subject: Important News About Your Bank Account
> Date: Sat, 24 Jan 2004 09:32:39 -0400 (EST)
> [snip]
> As a result Department Of Homeland Security Director Tom Ridge has advised
> the Federal Deposit Insurance Corporation to suspend all deposit insurance on
> your account until such time as we can verify your identity and your account
> information.

> Please verify through our IDVerify below. This information will be checked
> against a federal government database for identity verification. This only
> takes up to a minute and when we have verified your identity you will be
> notified of said verification and all suspensions of insurance on your
> account will be lifted.

  (the link behind the text was^A@
     [ORIGINAL triggered virus checkers.  BEWARE!  PGN]

> Failure to use IDVerify below will cause all insurance for your account
> to be terminated and all records of your account history will be sent to the
> Federal Bureau of Investigation in Washington D.C. for analysis and
> verification.

Avishai Wool, Ph.D.   +972-3-640-6316

Another wireless risk

<Chris Meadows aka Robotech_Master <>>
Mon, 26 Jan 2004 16:13:37 -0600

The other day I was in the position of needing to print out my credit card
site's invoice display.  Since I don't have a fully functional printer at
home, and I needed to make a photocopy anyway, I decided to take my Mac
Powerbook down to Kinko's and print it off there.

The problem was, when I plugged the Powerbook into their Ethernet link
(called a "Macintosh link" for some reason by their onsite
documentation...never mind that any computer with an Ethernet port could use
it), I couldn't reach the Internet.  (Nor could I see any printers in my
application...and the printer driver disk the Kinko's clerk helpfully
offered didn't help, because it only had drivers for OS 9, not OS X.)
However, the fellow who'd just vacated the laptop station had been using
wireless, and he said that should work.  And I did a quick scan, found an
open wireless router labelled "linksys," (the way they didn't even bother to
change the default name should have warned me, I suppose...but given the
general lack of computer adroitness I had observed in the staff, that
carelessness seemed to fit right in) with a Lexmark printer on it, and
Internet I called up the invoice and hit print, then asked the
Kinko's clerk where that particular printer was.

Longtime RISKS readers should be able to guess what came next.  "But we
don't have a wireless network...and we don't have any Lexmark printers
either."  Further research indicated that the wireless router was hooked
into a Bellsouth DSL connection, presumably someone's nearby home or
business.  So I had just printed my credit card invoice to some total
stranger's printer...and had no way even to find out where it was so I could
get it back.  Fortunately, the invoice didn't contain any *truly* sensitive
information, such as my SSN or account number (beyond "ends with ....").
And I was closing that account anyway.

The risk here is kind of the inverse of the "usual" risk associated with a
wireless system...instead of "you never know who might be using your
network," it's "you never know whose network you might be using."  The
combination of an open wireless network and a location where you would
expect there to be one can easily enough confuse you into conflating the two.

Hotel reservation system easily confused

<"Richard S. Russell" <>>
Fri, 23 Jan 2004 23:33:08 -0600

Our science-fiction group in Madison, Wisconsin, runs the world's only
feminist-oriented SF convention, WisCon. Every year we hold it in the same
hotel. Our publicity is required to say "Mention WisCon 2004 when making
reservations to get the group rate.". I asked why the "2004" part was
necessary and was told that, because of the hotel's automated reservation
system, the rates for WisCons 2002, 2003, 2005, 2006, etc.  are also in the
computer, and the reservation-takers apparently can't figure out from the
dates of the reservation (2004 May 28-31, if you're interested) which one
you're signing up for. Therefore, instead of building in a single central
error-checking process, they rely on a distributed network of hundreds of
naive human beings to each individually get it right -- assuming that the
convention committee has done ITS part and diligently included the year
every time it mentions the name of the convention.

Richard S. Russell, 2642 Kendall Av., Madison WI 53705-3736  1-608-233-5640

Browsers, online forms, rendering and opt-in marketing

<"Alistair McDonald" <>>
Wed, 21 Jan 2004 08:32:18 -0000 (GMT)

I completed a _long_ online for for the UK's DTI on information security
breaches ( if you
want to complete it. This took me over 10 minutes. At the end, there were
the usual options to untick the boxes and not have your information shared,
and so on. There was, however, a strange twist.

My browser (Opera 7.23 for Linux) failed to display the end of the page.  It
simply stopped halfway through one of the personal information lines.

I didn't want to fill in the form again using another browser, so I had to
view the source code and use the keyboard to navigate through the various
tick boxes.

Then, I had another thought: the form had questions like "If you do not wish
to receive information ... please untick here." The boxes were unchecked by
default - should I check them or leave them unchecked? The message indicated
that I should take an action - i.e. toggle the boxes, but perhaps they
should have been on by default, but my browser failed to do this, perhaps
due to non-standard HTML or script. By leaving them unticked, I'd be OK -
but no, I had to take an action.

However, it occurred to me that both of these errors could be used in a
deliberate way to confuse users and collect more personal information for
opt-in marketing.

By the way, the form was way too long. It was 266K, had nearly 7000 source
lines, and 224 input controls (including any hidden ones). It should have
been broken up into about a dozen smaller pages with Next and Prev buttons
to navigate between them. Of course, that would be more difficult to code.:-)

Drunk unlocks police car with own key

<Max <>>
Tue, 13 Jan 2004 20:57:58 -0800

Shinichi Kiyono, 32, was arrested on suspicion of car theft after he
reportedly mistakenly unlocked a police investigation vehicle with his own
car key while drunk, then drove to an empty lot and fell asleep in the
vehicle.  He turned himself in when he woke up in a Nissan that was not his,
although it was the same make and color.  Nissan says it makes more than
20,000 types of keys.

A spokesman for Nissan Motor Co., the maker of the vehicles, said the firm
produced more than 20,000 types of keys for its vehicles, and that it was
almost impossible for separate keys to be used in different cars, even if
they were the same model, but that it was not impossible for keys to very
occasionally fit other cars.  Prosecutors subsequently decided not to
indict him.

[Source: PGN-ed from the Japanese daily *Mainichi Shimbun*, 18 Dec 2003:
See also: "Drunk who unlocked police car with own key escapes prosecution":

  [I would be interested to see how many other police cars this guy's key
  can open.  Max]
    [Count the number of Nissans on the road, and divide by 20,000, to
    get a very rough estimate.  PGN]

Re: Happy 2**30'th birthday, time_t! (RISKS-23.12)

<Steve Summit <>>
Thu, 22 Jan 2004 18:40:14 -0500

Paul Eggert wondered how many time-related problems there might be in 2038
even though most machines by then will presumably be capable of using 64
bits.  I'm afraid the answer is: quite a lot.  Even if every CPU and OS is
using 64-bit time_t's by then, I expect there will still countless instances
of 32-bit time representations lying around on disk, baked into binary data
file formats which still reflect the original 32-bit size.  Upgrading CPU's
and recompiling programs will not, of course, automatically update all of
the terabytes worth of data files written and maintained by prior versions
of the programs.  (In other words, it's all too likely that 2038 will be to
Unix as Y2K was to COBOL.)

For those who use binary data files, a nice exercise is to write a pair of
functions for reading and writing between in-memory time_t values (however
big they happen to be today), and 6- or 8-byte on-disk values, making sure
that the functions are implemented such that they work without change when
compiled on a system with 32-bit time_t's, or recompiled for a machine with
64-bit time_t's.  (Of course, decoupling data file representations from
implementation-defined in-memory representations is almost always a good
idea; this is merely a timely example.)

Re: Suing the customers (Scrivner, RISKS-23.12)

<Paul Robinson <>>
Sat, 24 Jan 2004 19:43:32 GMT

> Rockwell, is suing a law firm that is currently suing Rockwell's customers.
> [...]

I think Rockwell doesn't have a leg to stand on.  A patent gives the holder
of it the right to prevent others from making, using or selling a patented
device.  Anyone in the chain of persons not having a license from the patent
holder can be sued.  If you purchase an ACME refrigerator from Pat's
Appliance Store, and it turns out that the ACME refrigerator has a
patent-violating component in the in-door ice maker, the owner of that
patent can sue ACME, they can sue Pat's Appliance store, and they can sue
you.  All three of you are jointly and severally violating the patent
holder's rights under the law and they can sue any or all of you.  Usually
the manufacturer is the only one who is sued but in theory anyone who
doesn't have a license, either directly or indirectly, is an infringer and
can be sued.

There was an incident a few years ago, when the manufacturer for the
electronic fare collection system implemented in the Washington Metrorail
system used some components that violated a patent (because the manufacturer
didn't have a license.)  The patent holder chose to sue the Washington
Metropolitan Area Transit Authority (WMATA) instead of suing the
manufacturer.  WMATA simply chose to settle by purchasing a license from the
patent holder.  I don't know if the transit authority ever got the extra
cost back from its supplier.

This is the exact same situation the RIAA is dealing with in the case of
people who are allegedly swapping songs over peer-to-peer networks.  (The
RIAA's real agenda is obviously control, not money, but suing people to
scare others is a fairly effective way to influence behavior.  If the RIAA
were interested in money, they would have taken up the offer of a billion
dollars in payments from licensing fees through sale of use of the service,
and Napster - the original one - would still be operating.)

The point is, even if the patent holder is wrong about their product being
infringed, legally they may choose to target the manufacturer's customers.
Some of them could conceivably counter sue if the company is intentionally
misusing the legal process but that's an iffy proposition, as some people
who tried to sue DirectTV over it's efforts to squeeze money out of anyone
who purchased a smartcard programmer from certain sites that sold devices
that allegedly could allow someone to obtain DirectTV's service, whether or
not the person actually did or could have used the programmer to unlawfully
obtain their signals, discovered.  A court found the attempts by DirectTV to
demand (enormous amounts of) money for alleged signal theft (whether or not
any actually happened) in place of filing suit was a legitimate action by
DirectTV, and ordered them to pay its costs to defend the case.

N.B. To prevent *me* from being sued in case I have named someone who really
exists, the name "ACME" and "Pat's Appliance Store" are fictional examples
not intended to represent any real-life company or organization.  :) But I'm
not worried anyway because I don't have any money and am unlikely to be
sued.  :(

>   In Rockwell Automation Inc. v. Schneider Automation Inc., 02-01195,
>   Rockwell says its technology is not covered by the Solaia patent, and
>   rather than battling that issue out in court, Niro Scavone and its clients
>   have sought to "'shakedown' manufacturers through threats of potential
>   business interruption or catastrophic damages."

It's not a 'shakedown' if you use the courts.  If I threaten you if you
don't pay me for something, that's extortion and a crime.  If I threaten to
sue you if you don't settle, that's legal.  If I just sue you anyway,
whether I have a case or not, that's also legal.  As many people are
relatively upset over in the case of SCO and it's claims it has some rights
over Linux due to alleged infringement.  It may be relatively slimy but it's
by the book legal, unfortunately.

  ["If the lessons of history teach us anything it is that nobody learns
  the lessons that history teaches us."]

Re: Lie-detector glasses, 90% accurate? (Holzworth, RISKS-23.14)

<Ron Bean <>>
Tue, 27 Jan 2004 22:01:04 -0600

> It may not be long before you hear airport security screeners ask, "Do you
> plan on hijacking this plane?"

I'd be highly tempted to reply "No, do you plan to stop beating your wife?"

Reading the rest of the article, it sounds like it's detecting people's
emotional "hot buttons", rather than lies per se. They talk about using it
as a "love detector" also...  (there should be a joke about lonely airport
screeners in there somewhere, but I won't attempt it).

The article also says:

  "The technology delivers not only a true/false reading, but a range of
  high-level parameters, such as "thinking level," which measures how much
  as subject has thought about an answer they give, and "SOS level," which
  assesses how badly a person doesn't want to talk about a subject."

I bet a good actor could reverse-engineer this, given enough
time with the machine.

> The company said that a state police agency in the Midwest found the lie
> detector 89 percent accurate, compared with 83 percent for a traditional
> polygraph.

What's the rate for false positives vs false negatives?
  [See the next item!  PGN]

Re: Lie-detector glasses, 90% accurate? (Holzworth, RISKS-23.14)

<"Peter B. Ladkin" <>>
Wed, 28 Jan 2004 11:31:14 +0100

Steve Holzworth reported that the manufacturer said that "a state police
agency in the Midwest" had found "lie detector glasses" to be "89%

That figure doesn't tell us anything about the usefulness of the glasses. To
obtain useful information, one needs to categorise errors as false positives
(that you are identified as lying when you are in fact telling the truth. I
shall call these Type 1 errors) and false negatives (that you are identified
as truth-telling when you are in fact lying. I shall call these Type 2
errors), and one needs to know the background rate of
truth-telling/lying. The company spokesman did not offer the classification
into types, and I doubt that he or the "state police agency" had any
reliable information about the background rate of lying.

To illustrate, let us interpret the "89% accuracy" statement as meaning that
the instrument is in error in 1 out of 10 uses. I consider three cases with
a 1 in 10 error rate.

Case 1: The background rate is also 1 in 10, all errors are Type 2, and the
instrument identifies no one as lying. Then Steve has zero chance of being
falsely accused.

Case 2: All errors are Type 1, and the background rate of lying is zero.
Then Steve has a 1 in 10 chance of being falsely accused.

Case 3: Errors are evenly split between Type 1 and Type 2, and the
background rate of lying is 1 in 2. Then Steve has a 1 in 20 chance of being
falsely accused. More worryingly, if he were to be intent on hijacking a
commercial aircraft, he also stands a 1 in 20 chance of passing the test
(Type 2 errors are 1 in 20)!

Now, consider what it would take to establish the background rate of
lying. Cut to the chase: it is difficult to impossible in serious use.

More specifically: This rate is likely dependent on the community, as well
as the selection procedure for testing, and it is also dependent on the
social importance of the proposition against which truth-telling/lying is
assessed. If everyone thinks lying is socially inappropriate, and you ask
them if they had exceeded the speed limit sometime in the last year, and all
know you have no possibility of enforcing sanctions given a positive answer,
then you are likely to obtain a very high rate of truth-telling, maybe even
perfect. On the other hand, if you sample a community in which more
importance is attached to getting off the hook than it is to whether one
tells the truth, but in which ceteris paribus truth-telling is preferred to
lying, and you ask people whether they have committed specific unsolved
murders, and sanctions are rigorously enforced, then everyone may well
answer "no" to each question. In this case, the background error rate is
identical to the unsolved-murder rate.

If you are Hercule Poirot, and you know the murderer acted alone, then you
know this rate (for, in his mysteries, there is a closed society, and
everyone professes innocence at first). Otherwise, one would have to perform
controlled experiment. Performing a controlled experiment is ipso facto
selecting one very specific value of community parameter, and is not
obviously a guide to communities which differ from that one.

In short, in case it wasn't obvious anyway, the company spokesman is BSing,
as are most people who claim to have measured the accuracy of lie-"detector"

Peter B. Ladkin, University of Bielefeld,

REVIEW: "Biometrics", Woodward/Orlans/Higgins

<Rob Slade <>>
Fri, 30 Jan 2004 08:16:43 -0800

BKBIOMTC.RVW   20031204

"Biometrics", John D. Woodward/Nicholas M. Orlans/Peter T. Higgins,
2003, 0-07-222227-1, U$49.99/C$74.95
%A   John D. Woodward
%A   Nicholas M. Orlans
%A   Peter T. Higgins
%C   300 Water Street, Whitby, Ontario   L1N 9B6
%D   2003
%G   0-07-222227-1
%I   McGraw-Hill Ryerson/Osborne
%O   U$49.99/C$74.95 905-430-5000 +1-800-565-5758 fax: 905-430-5020
%P   432 p.
%T   "Biometrics"

The book is intended for both students and professionals, covering all
of the aspects and uses of biometrics.  The chapters are written by a
number of contributing authors.  For example, Richard E. Smith, author
of "Authentication" (cf. BKAUTHNT.RVW) wrote the introduction found in
chapter one.  It is an excellent precis of the uses of, and
requirements for, authentication, paying particular attention to the
use, strengths, and weaknesses of biometrics.  The functional aspects
of biometric assessment; feature extraction, storage, error rates, and
so forth; are covered well in chapter two.  (There is a rather odd
confusion of genetic and phenotypic sources of biometrics: aside from
behavioural measures and DNA testing itself, almost all biometrics are
expressed characteristics, and therefore phenotypic.)

Part two deals with types of biometrics.  Chapter four provides
fascinating details on the history, technology, storage, indexing, and
searching of fingerprint records, and a brief mention of hand
geometry.  After the wealth of technicalities about fingerprints, the
very basic explanations of enrollment of face and voice recognition
are disappointing.  The material on iris and retina scanning, in
chapter five, is slightly better, but signature and keystroke dynamics
again get minimal coverage in chapter six.  Eleven of the more
esoteric biometrics are briefly described in chapter seven, ranging
from standards such as DNA testing to odd entries like sweat pore
distribution or body odour.

Part three looks at various aspects or factors to consider in
implementing biometrics.  Chapter eight looks at the question of
"liveness" testing.  (This is the biometrics topic beloved of students
the world over: "What if you cut off the guy's finger and used that?"
Students tend to be rather gruesome creatures.)  Most of chapter nine
is devoted to a guide for contracting out, or questions to ask
contractors or vendors.  Various standards bodies are described in
chapter ten.  Chapter eleven talks about issues involved in testing of
biometric systems.

Part four deals with privacy, policies, and legal issues.  Chapter
twelve examines both the threats and the benefits that biometrics
holds for privacy.  There is a detailed and interesting look at
(mostly US) law and decisions relating to privacy, and the
implications for biometric applications, in chapter thirteen.  Chapter
fourteen does have brief case studies of the use of biometrics at the
Super Bowl and in Virginia Beach, but concentrates on the legal
issues.  Chapter fifteen deals with the American digital signature
law, and the potential relation to the inclusion of biometrics in the
process.  Some material is repeated from earlier chapters.

Part five reviews selected biometrics programs.  Chapter sixteen
covers government and military programs, most related to law
enforcement.  Searching the FBI files of civil (or non-criminal)
fingerprint files, in chapter seventeen, reiterates a fair amount of
content from chapter four.  Private sector programs, in chapter
eighteen, are primarily concerned with face recognition in casinos or
a variety of systems for banks, but others are mentioned.  Chapter
nineteen presents a very detailed and thoughtful analysis of the
possibilities for a national identity card.

Because this book is essentially a collection of standalone essays by
a variety of authors, there is a great deal of overlap and duplication
of material, and at times this repetition becomes annoying.  This is,
however, the most useful and informative work on biometrics that I
have reviewed to date, and the analysis, in particular, is
comprehensive and even-handed.  I would recommend this as both a
serviceable introduction to anyone who must work with biometrics, and
as a guide to the controversies surrounding them.

copyright Robert M. Slade, 2003   BKBIOMTC.RVW   20031204    or

Please report problems with the web pages to the maintainer