The RISKS Digest
Volume 16 Issue 50

Tuesday, 25th October 1994

Forum on Risks to the Public in Computers and Related Systems

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

Please try the URL privacy information feature enabled by clicking the flashlight icon above. This will reveal two icons after each link the body of the digest. The shield takes you to a breakdown of Terms of Service for the site - however only a small number of sites are covered at the moment. The flashlight take you to an analysis of the various trackers etc. that the linked site delivers. Please let the website maintainer know if you find this useful or not. As a RISKS reader, you will probably not be surprised by what is revealed…


Another way for computer failure to delay trains
Mark Brader
Bank Robber Foiled by Security Screen
Paul Gillingwater
Re: Ignorance in Computer Science
Roy Maxion
Re: Don Norman and the Power switch
William Hugh Murray
Re: deadly risk
Scot E. Wilcoxon
Re: Cell Phone Fraud Countermeasures
Bill Hensley
Re: Badly designed interface in MS Mail
Russell Stewart
Re: Inadvertent postal forwarding
Kent J Quirk
Re: Data security in Iceland
Curtis Jackson
Mailing lists risk critical-mass spamming
Benjamin A Ostrowsky?
Sears takes digitized signature for CC transactions
Steve Holzworth
Lauren Weinstein
Michael D. Sullivan
Tim Duncan
B.M. Cook
J. Eric Townsend
Michael Jampel
Mark Stalzer

Info on RISKS (comp.risks)

Another way for computer failure to delay trains

Mark Brader <>
Tue, 25 Oct 1994 18:15:21 -0400
Operational notices for the US passenger rail service Amtrak are now being
posted to rec.railroad, and I spotted this item in one of them:

  Termination of Train 352(19) at Detroit, MI

  1045pm, ET train 352(19) diesel 359, CC 9649, 3 cars, arrived
  Detroit on time.  Train was terminated due to no engineer was
  available to handle train to Toledo due to a Crew Management
  System failure.  Regular man had called in at 245pm, ET, and
  advised was having automobile trouble and would not make train and
  marked off.  Crew Management advised was having computer failure
  but engineer had marked off prior to the computer system outage.
  Investigation to be scheduled.  Passengers were handled via
  alternate transportation to Toledo.  Equipment was deadheaded to
  Toledo when relief engineer was called.
  Delay:  352(19)  term.  passengers delayed 1'05"

Notations: MI = Michigan, ET = Eastern Time, 352(19) = train number 352
on the 19th of the month, engineer = driver, 1'05" = 1 hour 5 minutes.

Mark Brader  SoftQuad Inc., Toronto

   [Presumably, the relief engineer was not a grateful deadhead.  PGN]

Bank Robber Foiled by Security Screen

Paul Gillingwater <>
Wed, 26 Oct 1994 00:20:16 +0100
A New Zealand newspaper, "Waikato Times", for Sept. 2, 1994, reports a
coroner's court finding that a bank robber in Sydney (Australia) died after
being trapped for more than 15 minutes by a 250-kg steel security screen,
which rose to the ceiling and trapped him by the throat.

The screen had been activated by staff in July 1991, whilst the robber was
crawling across the counter.  The robber, Steven Kovacs (age 37), died of

The risk?  When designing control systems for steel bullet-proof security
screens that actuate in less than half a second, consider what happens if
someone is in the way! (Paul Gillingwater) :: Home Office in Vienna, Austria

    [I suppose this would be more computer-related if it were
    triggered automagically, but it is still certainly RISKful.  PGN]

Re: Ignorance in Computer Science (Goland, RISKS-16.49)

Roy Maxion <maxion@K.GP.CS.CMU.EDU>
Tue, 25 Oct 94 19:11:44 EDT
> ... we have no such thing as 'human engineering classes', 'interface design
> classes', or 'software management' classes.  ...

I hope you will be pleased to note that here at Carnegie Mellon I am teaching
just such a course.  It's called Dependable System Design, and it's mission is
to teach students how to design/implement/evaluate dependable systems so that
we see fewer problems of the kind recently described in RISKS.  Excerpts from
the course description follow:

  System dependability has been a major concern since the dawn of the
  electronic computer age.  Dependability, taken in its broadest context,
  includes metrics such as reliability, availability, testability,
  maintainability, usability, mean time to failure, etc.
  The goal of dependable system design is to ensure that the system continues
  to deliver its specified service to the client even in the face of error on
  the part of the user, the system or the environment.
  User interfaces, while just as critical for dependability as reliable
  hardware and software, have received little attention from the dependability
  and fault-tolerance community. Because the philosophy of dependable system
  design crosses the boundaries of hardware, software and user interfaces, a
  major portion (but not all) of the course will focus on the most recently
  critical aspect of dependable systems: user interfaces and user-system
        Upon completion of this course, the student should be able to: acquire
  system requirements, develop system specifications embodied in one or more
  dependability metrics; critique an initial design; identify weak portions of
  a design; suggest a variety of alternatives to improve dependability; use
  contemporary techniques to evaluate, compare and contrast alternative
  designs, conduct system-evaluation experiments, and produce balanced,
  cost-effective system designs.
  Topics include: Theory and practice of traditional dependability; robust
  benchmarks and fault injection; dependable user interfaces; experimental
  design; concepts of reliability and validity; system evaluation techniques;
  requirements acquisition; human error and reliability - causes, effects and
  mitigation; system monitoring, diagnosis and data analysis; task analysis.

  Dependability Motto:
    Anyone can build something that works when it works.
    But it's how it works when it doesn't work that counts.

Roy Maxion  Computer Science Dept.  Carnegie Mellon University

      [And WHO'S counting?  PGN]

Don Norman and the Power switch

William Hugh Murray <>
Tue, 25 Oct 94 15:21 EST
Don's tale of woe reminded me of my colleague who was assigned the task of
standardizing keyboards across IBM.

And they wonder why we security people cannot decide on safe defaults.

William Hugh Murray, New Canaan, Connecticut

  [DeFault is easy to assign.  But don't
  Pin the Tale on the Don Quixote.
  BTW, Don Norman says he has NEVER had so many responses to any of his
  previous RISKS submissions as he had on the power-switch posting.
  He must have pushed a lot of your buttons.  PGN]

Re: deadly risk

Sun, 23 Oct 94 17:38 CDT
>... electrocauterizer had apparently corrupted the software of her micro-
>processor-controlled pacemaker ... I believe the risk here is pretty obvious.

Yes, it is obvious that the woman might be in danger whenever she is near loud
music.  I hope they don't really use AUDIO tones.

A less deadly risk is assuming that experienced RISKS readers can only find
one risk.

Scot E. Wilcoxon      +1 612-936-0118

Re: Cell Phone Fraud Countermeasures

TRW <>
Sun, 23 Oct 1994 00:48:49 -0500 (CDT)
A recent post described a TRW effort to use technical means of defeating
cellphone fraud, and was described as voice print recognition.  FYI, the
actual method stores the electronic signature of the telephone instrument
transmitter.  A cloned phone will have a different transmitter signature and
calls from that phone will be blocked.  This system has been described in some
detail in Communications Week, and several other trade journals (that's where
*I* get my info :)

Bill Hensley, TRW Oklahoma City Engineering Office  405.732.3433 (v)  405.737.2043 (f)

Re: Badly designed interface in MS Mail (Astoorian, RISKS-16.46)

Russell Stewart <>
24 Oct 1994 19:08:14 GMT
>...  Since "New Password" and "Re-Enter New Password" were both
>empty (and thus matched each other), pressing OK set her password to the
>null string without her realizing what had happened.

This is extraordinarily stupid on Microsoft's part.

I have only been programming for about 6 months now (and only with Visual
Basic), but one of my fundamental rules is to NOT allow the user to leave
a dialog box until he or she has supplied all required information (or,
in the case of something ambiguous, like an empty password, at least ASK
the person before moving on).

If anyone finds the programmer responsible for that mistake, he should be
taken out and shot (and then dug up and shot again).

Re: Inadvertent postal forwarding (Bove, RISKS-16.48)

Kent J Quirk <>
Sat, 22 Oct 1994 00:13:52 -0400
In RISKS-16.48, V. Michael Bove Jr. writes about a post-office forwarding
problem.  We had a similar problem once.  My wife once ran a small business
called Engineering Services Group.  When we moved, she forwarded mail to our
new address.  One of the mailing lists she was on was at Hewlett-Packard (HP).
 HP got the change of address notice, and (apparently) maniacally changed
*every* occurrence of Engineering Services Group in their list to our new
address.  As you might imagine, several companies have an ESG as an internal
group — something like this:

    H. Dilbert Gweep
    Graphic Doodler
    Engineering Services Group
    Habeas Corp.
    Anytown, OH, 00666

All 25 of these people suddenly stopped getting their 6-color HP spreads on
90-pound stock.  All 25 copies suddenly started arriving in our mailbox with
disconcerting regularity, each one with a different name.

The hard part was getting HP to turn the bloody thing off.  The 800 numbers
printed on the literature were useless.  I finally found a person named on the
masthead of some "magazine", and called her directly.  After several
cross-country phone calls, she eventually managed to reach someone who could
(and did) make it stop.

The risks?  Besides the danger of herniated homeowners and mail carriers,
there are 25 people out there who are no longer on a mailing list they might
have actually wanted to be on.  I doubt HP kept their old addresses lying
around.  Imagine if her company had been named "Engineering Department", or
worse yet, "Craig Shergold".


(For the urban-legend-impaired, Craig Shergold is the now-grown English boy
who no longer wishes to receive a world-record number of postcards.)

Kent Quirk  7 MacLeod Lane  Acton, MA 01720  508-263-8344

Re: Data security in Iceland (Haynes, RISKS-16.47)

Sun, 23 Oct 1994 00:38:33 GMT
}...  If only we could get the story to be spread and repeated as well as the
}Craig Shergold plea and the gold star tattoo urban legend!

If you could, it might make a difference to the small business people, but
might not make a difference to an individual.

Back in the "good old days", hard drives were small enough that backup to
floppies was feasible on a regular basis, and it was therefore cheap.

Today, I have a 1GB external disk on my Mac that I use for all my work,
including lots of transitory data that I don't care about, and I backup the
things I do care about to my internal 230MB drive. If my computer is stolen, I
will certainly lose everything.

Am I naive? Stubborn? Stupid? Well, possibly. But my motivation in this case
is cost. Unlike IBM-compatible PCs, sizeable removable media backup systems
for Macintoshes are extremely costly. It is *much* cheaper to simply keep
buying bigger and bigger hard drives, and use the older smaller one for backup
and the newer larger (and faster) one for main use.

I am choosing not to "pay for full coverage insurance" for cost reasons; I am
protected if my main hard disk dies an untimely death, but not if all of my
computer equipment is stolen.

Consider it a RISK of one technology far outstripping another technology on
which the first is dependent. Non-removable media cost per MB, speed, and even
reliability have far out- paced that of removeable media.

Curtis Jackson

Mailing lists risk critical-mass spamming

Sat, 22 Oct 1994 00:37:13 -0400
     At what point is it worth a lawyer's time to spam the net?  A certain
size of audience must be reached, and easily.  Public mailing lists risk
presenting an easy way to spam.
     Many lists accept posts from non-subscribers.  One need only know
that the list exists and where to send contributions.  It would be a simple
task to poll ten thousand LISTSERVs for a list of lists.  Having compiled
such a listing, one could then send one's advertisement into several
hundreds of thousands of mailboxes.
     A more pointless exercise, and one which would cause even greater
havoc, is to set up a LISTSERV list on one's own system which would link
every list to every other list.  It takes no imagination at all to see
how quickly this would create a logjam.

Sears takes digitized signature for CC transactions

Steve Holzworth <>
Tue, 25 Oct 1994 07:55:55 -0400 (EDT)
A few days ago, I was in the local Sears store (Cary, N.C.). I was purchasing
a few items and decided to charge them against a Visa card.  The clerk
dutifully rang in the charges. While she was doing that, I noticed that a
small digitizing tablet was attached to the register.  I had a feeling I new
what was coming next...

She pulled the receipt out of the register, slid it under a hold-down on the
tablet, and instructed me to sign within the little plastic guide.  I refused,
stating that I would gladly sign the receipt, but not on the tablet.
Apparently, I was the first to refuse :-) She had to look up a magic phone
number, ominously entitled "Loss Prevention", to determine what to do. The
person on the other end OK'd my signing without having my signature digitized.

Some might argue that my refusal was meaningless because Sears could always
scan my signature from the receipt. Maybe so, but nothing says I have to make
it easy for them. Small encroachments eventually lead to large losses.

Steve Holzworth  SAS/Macintosh Development Team  SAS Institute  Cary, N.C.

CNID (numbers vs. names)

Lauren Weinstein <>
Fri, 21 Oct 94 20:46 PDT
The assertion has been made that perhaps calling number ID would be more
acceptable if something other than numbers were sent.  In fact, that is
already part of the CNID spec (calling name delivery), though it is not
implemented in all CNID systems.  In most implementations I've seen *both*
name and number are delivered--which rather kills the idea of delivering "who
is calling" info rather than "where are they calling from" info.

There isn't much enthusiasm among the commercial end-users for name-only
delivery.  The reason is obvious--you can't collect phone numbers for return
sales calls when you're just handed a name.  While the telcos have touted CNID
as being for individuals, the real focus has always been on commercial use as
the real money maker.

At this point, the CNID controversy boils down to the seemingly simple
question, "Since it is mandated that everyone must have per-call blocking
available, why can't per-line blocking be made available?"  Once again, I
think the reason is obvious.  Most people aren't on PBX systems that are
easily programmed to dial the three digit blocking code on each call, nor will
most people want to spend money for add-on gadgets to do this.  So the hope of
the telcos is that most people will forget or won't bother to use the per-call
code most of the time.  If per-line blocking were available (and surveys show
most subscribers would want it--people want to know who's calling, but don't
want other people to always know *their* number!) then CNID's commercial value
would be significantly lowered.


Re: CNID (Preece, RISKS-16.47, Klossner, RISKS-16.48)

Michael D. Sullivan <>
22 Oct 1994 02:44:36 -0400
Andrew Klossner objects to the proposal to assign each phone a second unique
number solely for CNID; This would be accessible only through an 800 number.
Klossner claims this would be impractical because "Almost all area codes in
North America are more than half populated. In order to double the number of
phone numbers, most area codes would have to be split."  This does not present
a major obstacle, however.  The pseudo-number would not have to be a valid
phone number.  [This was also pointed out in a subsequent message by Scott
Preece himself.  PGN]  Such a number could fit within the existing CNID
transmission system if it utilized the same number of digits as a valid US
number (10 digits), but there are no valid 10-digit numbers beginning with 0
or 1 (20% of the numberspace), no valid 10-digit numbers with 0 or 1 as the
fourth digit (20% of the numberspace, overlooking overlap with the foregoing),
no valid 10-digit numbers with 11 as the second and third digits (reserved for
911, 411, and other service codes) (1% of the numberspace), no valid numbers
with 11 as the fifth and sixth digits (same reason) (1% of the numberspace),
no valid numbers with 00 as the fifth and sixth digits (to avoid confusion
with 800, 900, etc.) (1%), no valid numbers with 200, 300, 400, or 600 as the
leading digits, etc.  Over 40% of the numberspace is reserved, and the
remaining 60% is far from fully occupied — only recently have a handful of
area codes with 0 or 1 in the middle been assigned (20% of the numberspace
remains relatively vacant), and likewise phone numbers with a 1 or 0 as the
fifth digit (i.e., in the middle of the NXX "exchange" group) have been
assigned only in major urban areas.  It is very unlikely that anywhere near
50% of the numberspace is actually assigned — keep in mind that there are 10
billion unique numbers available in a ten digit numberspace.

Clearly, then, it would be technically possible for the foreseeable future to
assign each telephone number in the North American Numbering Plan a unique
second number.  This number would not be usable as a telephone number, but it
would only work through an "800" number, which would decode it.

Moreover, there is no reason why such a non-dialable number should be uniquely
assigned to each number.  The telephone companies could use the non-dialable
subset of 10-digit numbers as a pool of codes for one-time use in each area
code.  The real dialable number is actually transmitted between central
offices, but is not transmitted to the end user.  If the "privacy" bit is
toggled on, the central office could store the real number and transmit a
non-dialable number from a one-time pad, and it would only be mapped to the
real number for a limited time, such as 48 hours, if keyed into the "800"
number associated with the same telephone company and dialed from the number
that was called (i.e., from the phone that received the non-dialable number).
There is no reason for a permanent association between the non-dialable number
and the real number; in fact, that would defeat the purpose of CNID and
unpublished numbers.  Calling the pseudonumber, through the 800 gateway, would
be a one-time deal.  There's no major technical obstacle to that.

Michael D. Sullivan | INTERNET E-MAIL TO:  |also: |
Washington, D.C.    | | |

   [Various astute messages from (Paul T. Keener),
      "john (j.m.) clarke" <>,
      George Swan <>,
   and others tend to make similar observations, and are omitted.  Please
   forgive me if I have oversimplified the points made by the omitted plethora
   of messages.  I think you would all be overwhelmed if I included
   everything.  PGN]

Re: CNID and Don Norman — CNID can be private (Wells, RISKS-16.48)

Tim Duncan <>
Sun, 23 Oct 1994 13:18:42 +0000 (GMT)
|> [Don Norman] suggests that other information should be sent instead of a
|> telephone number. ...

This is currently being tested by British Telecom in Edinburgh.  Subscribers
have been given the option of submitting their own identifying string (16
characters) instead of the default which is their name as it appears in the
phone book.

Tim Duncan, AI Applications Institute, Univ. of Edinburgh, 80 South Bridge,
Edinburgh EH1 1HN, Scotland, UK  Tel: +44 31 650 2747  Email:

Calling Number ID - in the UK

"B.M. Cook" <>
Tue, 25 Oct 1994 09:26:24 +0000 (GMT)
We're about to get calling number ID in the UK, too - from British
Telecommunication (BT) at least.  It is being promoted on the benefits to the
subscriber, e.g. "This service has already been available in North America for
some time, where it has had a major impact on malicious and fraudulent

The fact that lines for which information has been previously withheld
(ex-directory numbers) will send information is hard to spot in the

The features listed in DIGEST 16.46 appear to have implemented, but not well
publicised - I get the distinct feeling that the service will be of most
benefit to companies who want to create lists of 'phone numbers!

We will be getting -
1. Per-line blocking - FOC, just call and ask for it.
2. Per-line unblocking - the default unless you change it.
3. Per-call blocking - dial 141 before the number you want.
4. Per-call unblocking - dial 1470 before the number you want.

1. and 2. need a phone call but it costs nothing.
3. is advertised in the literature.

BUT 4. took some discovering!  I want to block outgoing ID by default and
allow it only on certain calls (e.g. friends, certainly not businesses who
will record it, sell the lists etc.) so per-line blocking with per-call
unblocking is the answer.

This option is not mentioned in the literature, I phoned the further
information freefone number to be told that what I wanted is not possible.
Rather than give up here, as I guess most people would, I asked them to record
my requirement so that if enough other people also wanted it they might think
about providing it.  They said they'd get someone to call me back which they
did the next day.  A very pleasant call from Edinburgh from someone who knew
rather more about the system (it had been on trial in Scotland before being
introduced elsewhere).  No problem, he said, we already have what you require,
just have the line blocked and put 1470 in front of calls you want to unblock -
I'll put the line block on for you. The next day we had confirmation, by mail,
that it had been done.

So, it looks as though we will be getting a decent service (it goes live on
November 5th) - but only if you know what you want and work a little to get
the information!

Bypassing CNID in an emergency

J. Eric Townsend <>
24 Oct 1994 05:05:25 GMT
I see a big RISK in "ignore all phone numbers I don't recognize".  At least
once every 6 - 12 months, I get an emergency call at my house from a friend or
family member.  The call usually comes from a pay phone, but has originated at
a business or a residential line as well.

By blocking "unknown numbers", I would have missed calls with the following
content in the past year.

- "best friend severly injured in a vehicle accident, come immediately"

- "we broke down coming back from a late nite concert, it's 0230 and
   we're stuck in the middle of a really bad section of town with no
   transportation.  can you come get us?"

- "my flight was canceled, so I caught one that arrives 3 hours earlier
   and at a different airport.  come get me before I'm bored to death!"

And, most importantly to *me*:

- "Hi, I just got your resume from a friend of yours, and we'd like
you to come in for an interview."

I ended up getting a great job from that one...

I don't see what is signifcantly gained by CNID that isn't gained by
using an answering machine to screen calls.  I do see what can be lost
by ignoring calls from numbers one does not recognize, however.

J. Eric Townsend  USA 415.335.7463 aka
work:      AT&T PersonaLink:

Caller Number Identification in the UK

Michael Jampel <>
Mon, 24 Oct 94 10:45 GMT
British Telecom is starting a CNID (Caller Number Identification) in the UK.
If you do nothing, your number will be sent out every time you make a call.
But if you dial 1471 (I think — anyway, some 4 digit code) before dialling
the number you want to call, then your number will not be sent out.

Or you can ask for your number _never_ to be sent out. But in this case it is
_not_ possible to dial, say, 1472 to have your number sent out occasionally.

So if you don't particularly want to tell life insurance companies your phone
number, and if you don't trust yourself not to forget to dial 1741
double-bucky, you are put in the situation where you can never send out your
number, which will usually not matter, until the day that it does matter.

I feel that BT are not serving the needs of their domestic customers here:
CNID is great if someone you know is getting dirty phone calls, but otherwise
its main function is to benefit commercial customers. But there is nothing we
can do about it.

Michael Jampel

Re: CNID and Don Norman — CNID can be private

Mon, 24 Oct 1994 09:03:41 +0800
It would not take the database miners of the world long to build a cross
reference between user selected call number ids and the real phone numbers.
To keep off the list, you would have to never let a company associate your
real phone number with your id. This would be hard to do if you use credit
cards, mail order outfits, etc.  Also, given the penchant for phone companies
to charge a few bucks for every trivial service change, like changing an id,
the cross reference would tend to remain up to date.

Although, I should add that if people conspire to choose the same id,
like PRIVACY, we can give the marketers a real headache.

  — Mark

Please report problems with the web pages to the maintainer