The RISKS Digest
Volume 17 Issue 71

Tuesday, 13th February 1996

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…


The measurement of risk: community measures vs scientific measures
Dave Shaw
Those fun-loving guys at LANL.GOV
Simson L. Garfinkel
More on WWW-Robot false hits...
Debora Weber-Wulff
Re: Risks of web robots
Cameron Simpson
Re: RISKS (...) of typing credit-card numbers
Olin Sibert
Mark Fisher
Subject: Leahy to introduce bill to repeal CDA!
Stanton McCandlish
Foreign `replies' cause anxiety
Timothy Mowchanuk
Correction: Train operators get permission to use manual backup
Jonathan Kamens
Re: Electronic Medical Records and Images
David Coburn
Allan Noordvyk
Tom Olin
Info on RISKS (comp.risks)

The measurement of risk: community measures vs scientific measures

Dave Shaw <>
Mon, 12 Feb 1996 17:20:42 +1100
This e-mail is a summary of, and quotes heavily from, an article entitled
"Outrage at the unknown" that appeared in the *Sydney Morning Herald* on 8
February 1996. The article was written by Simon Chapman, the associate
professor of Public Health and Community Medicine at the University of
Sydney, Australia. Chapman himself quotes Peter Sandman of Rutger's

Recently (late 1995 I think) Telstra (the public telecommunications company
of Australia) tried to install a new mobile-phone tower adjacent to a
kindergarten in Harbord (a suburb of Sydney). A group of residents, with
their young children carrying placards, chained themselves to Telstra's
fence, and called in the media complaining about (among other things) the
danger of electro-magnetic radiation (EMR).

Within two weeks, Telstra had lost face and switched off the installation. A
fortnight later the same battle erupted in another suburb, and it's happened
again in Harbord after the antennas were moved to a different site.

This continuing row is proving a feast for people in public health who study
the public's reaction to health risks. These "risks analysts" address questions
of how people might be motivated about a serious risk and calmed down about
risks that are trivial.

To a scientist, risk is the magnitude of the danger multiplied by the
probability of exposure. When the furor over the towers erupted, Telstra
called in the Australian Radiation Laboratory. Their EMR meters barely got
the needle off zero. It would be hard to find a serious scientist who would
call the towers a risk.

So why did the towers cause such outrage? Because this way of measuring risk
is not how ordinary (i.e., non-technical people) think. Chapman lists
Sandman's seven characteristics that determine an issue's "outrage valency"
in a community.

Voluntary vs coerced risks: if something is voluntary it will be much more
acceptable, eg. skiing while dangerous, is voluntary, and is not widely
dreaded.  A major theme of the Harbord protest is that Telstra (the "big
company") rode roughshod over the Harbord residents (the "little people").

Natural vs industrial risks: there are countless examples of communities
being upset by chemicals that have been "put there" by government/industry,
while not showing the slightest concern for the natural supply of these
chemicals.  For example, anti-fluoridationists are often shocked to find
that some areas contain naturally high levels of fluorine in the water.
Certainly, Telstra's towers are quintessentially industrial.

Familiar vs exotic risks: mobile phones are new and exotic and hence attract
outrage. The longstanding electricity substation, located at ground level
opposite the kindergarten, which emits comparatively huge amounts of EMR, is
familiar and attracts no outrage.

Not dreaded vs dreaded: diseases like cancer, AIDS, plague, tuberculosis
create a great deal more public concern than others, eg. heart disease. The
Telstra towers "might" cause cancer in children.

Chronic vs catastrophic: Thousands are killed on roads (as part of driving
accidents) but rarely in large groups. Plane accidents are much rarer,
cause fewer deaths, but because they can cause large fatalities, air travel
is much more widely feared than car travel. [Perhaps this explains (in part)
the number of articles regarding air safety in comp.risks ;-) ] Telstra's
towers "might" cut swathes into the future health of the local children.

Knowable vs not-knowable: much to the dismay of Telstra, the fact that the
Australian Radiation Laboratory could barely detect any EMR from the towers,
actually seemed to increase outrage (not decrease it).

Morally irrelevant vs morally relevant: communities tend to have
zero-tolerance for outrage-inducing problems. Telstra tried to favourably
compare the amount of EMR in TV sets, microwaves, sunlight and a fluorescent
lamp, but to no avail.

Well, that ends the summary, what follows are my thoughts.

The Telstra towers meet all of Sandman's characteristics. It would be no
surprise to him I suspect of the outrage they induced in the community.

The RISKS [sorry you had to wait so long for it :-( ] for those of us in the
scientific/industrial community is that we fail to understand how the
community measures risk. As Sandman says: "Experts focus on hazard and ignore
outrage. ... The public focuses on outrage and ignores hazard."

Understanding how the community measures risk should enable us to develop
more acceptable products or at least make their introduction easier. While
it is useful to have the public vet all products (to avoid the introduction
of "nasties"), technology has the potential to benefit everyone, and
unnecessary delay or fear, reduces that benefit.

So when we see cases like the Telstra towers we should not dismiss the
reaction as the fears of an ill-informed and untrained public, but rather
see it as the normal response of a community that measures risks according
to different criteria.

Hence, if Telstra could have found a better way of measuring the risk of
their towers (i.e., the Radiation Laboratory and their EMR meters), they may
have avoided publicity like the angry Harbord father, jabbing his fingers on
TV, at dour Telstra officials, and shouting: "There's no way I am going to
let my daughter or any of my kids go to that place and be exposed to that
sort of risk!"

David Shaw, Alcatel Australia Limited

Those fun-loving guys at LANL.GOV

Simson L. Garfinkel <>
Sun, 11 Feb 1996 09:12:18 -0500
Reading about LANL's "(Click here to initiate automated `seek-and-destroy'
against your site.)", I was reminding about a story that happened to a
friend of mine a few weeks ago.

Turns out that my friend was writing a web-walking robot, and it made the
mistake of walking into the LANL site. This robot was running at the end of
a 28.8K SLIP link, so it wasn't capable of issuing more than 1 request every
2-3 seconds.

Well, the folks at LANL have some sort of monitoring software, because they
noticed it immediately. What they did was they called up his Internet
service provider and said that he was attacking a federal interest computer,
and they threatened the ISP that unless they revealed the name and phone
number of my friend, LANL would take legal action against the ISP.

Those fun-loving guys at LANL then called up my friend and left the
following message on my friend's answering:


The folks at LANL then called my friend's Internet service provider and
threatened them with legal prosecution for violation of various computer
crime statutes, unless the ISP cut off my friend's Internet connection.

This is really scary --- the thought that some government official can call
up your ISP and, through a combination of threats and legal citations, have
somebody's Internet feed immediately terminated. What about due process of
law? What about innocent until proven guilty? What about having to go
through the mere formality of obtaining a court injunction before having
action such as this taken?

More on WWW-Robot false hits...

Debora Weber-Wulff <>
12 Feb 1996 14:02:09 GMT
A few weeks ago our WWW server was brought to its knees. We were being
inundated with thousands of URL requests for a student's home page. The page
didn't look that interesting, but we closed out the account and put out an
all-points search for the student in question and tried to figure out what
the entire world wanted from this student.  Theories varied from a viral
attack to wayward robots.

When he was dragged into the computer services center, he confessed to what
he had done: He had installed one of these nifty counters to see how many
times his pages was read. Since he had no hits other than himself, he
decided to include some good names on his page (a little racier than "sex 'n
drugs 'n rock 'n roll", but this is a family publication) and then he
registered his page with "some cyberporn list".  He did not remember which
one it was. So apparently all the robots in the world found a new site that
seemed to have racy new stuff in it; It was duly registered and there appear
to be quite a number of people that either automate the search for sex
pictures or check out what's new first thing in the morning; many are so
stupid and keep trying, even when the server tells them that this link is no
longer in operation.  It took days for things to calm down. Needless to say,
the student currently has his net account revoked...

The moral of the story: don't attract robots with false claims, there
are too many of them out there!

Debora Weber-Wulff, Technische Fachhochschule Berlin, FB Informatik,
Luxemburger Str. 10, 13353 Berlin, Germany

Re: Risks of web robots (Dellinger, RISKS-17.66,67)

Cameron Simpson <cameron@dap.CSIRO.AU>
Tue, 13 Feb 1996 16:36:23 +1100 (EDT)
There is a protocol called the Robot Exclusion Protocol designed explicitly
to prevent robots from traversing on-the-fly datasets. It solves exactly the
problem outlined above.

A moment's search through Yahoo reveals:
entitled "A Standard for Robot Exclusion".
    - Cameron Simpson, DoD#743

Re: RISKS (...) of typing credit-card numbers (Sibert, RISKS-17.69)

Olin Sibert <>
Mon, 12 Feb 96 15:44:33 EST
In response to messages sent to, but not included in RISKS, Bob Dolan (Xerox
Corp.) observes that countermeasures are a continuing battle:

   Date: Thu, 8 Feb 1996 12:09:18 PST
   Subject: RISKS (and lack thereof) of typing credit-card numbers

   It appears to me that no matter what methods are employed to detect
   these types of intruder programs, new approaches will always develop.
   It is consistent with the cracker mentality.

Fortunately, the pace of development in malicious software seems to be
fairly slow.  Once an initial strain is reasonably well neutralized,
later more virulent versions don't seem to propagate as fast.

Bob Dolan also identifies a fraud prevention problem:

   The flaw in this system lies in the fact that anyone can use a credit
   card to order merchandise and have it shipped to any address.  If the
   credit-card number implied a shipping address that matched the
   billing address, the numbers would lose their value to the crackers
   who steal them.  I realize this does not work for all users.
   However, anyone who desires secure transactions, would certainly want
   to use this type of credit card.

Actually, the credit-card companies do make fair use of delivery address
in guarding against fraud; some reader here no doubt has had a friendly
call from American Express asking whether they really meant to order 24
leather coats for delivery to downtown L.A.  However, with digital
goods, this is a more serious problem, as there's no physical "delivery
address" that can be meaningfully validated.  As with existing fraud
detection schemes, there will always be a battle of wits going on, and
some cost of fraud to be absorbed in the costs of doing business.
Delivery address validation for electronically ordered physical goods
is something the credit-card companies have well in hand today.

Charlie Abzug questions the feasibility of untraceable delivery:

   Date: Wed, 07 Feb 1996 22:06:37 EST
   From: Charles Abzug <>
   Subject: Capturing keystrokes of credit-card numbers

   I would like to add in my two-cents' worth:  The claim that the
   credit-card numbers would be sent out "sub rosa" and thereafter would
   be untraceable does not bear up to close examination.  Once the
   Trojan Horse software is discovered on someone's computer, it can be
   decompiled or disassembled to yield the full information on the
   E-mail address or other scheme by which the information is sent out
   of the compromised host computer to some unsuspected destination.

The problem is not in determining how the data is delivered, but who the
recipient might be.  True, you can find out what the mechanism is by
analysis and disassembly, but that doesn't tell you who actually gets the
messages.  Delivery to an anonymous recipient isn't hard--one need merely
take advantage of the Internet's broadcast media, such as mailing lists and
newsgroups.  Cryptographic encoding of the 60-odd bits that make up a credit
card number into, say, a Message-ID field in mail or a news posting would be
sufficient.  It doesn't even have to be a particular newsgroup: the
perpetrator can simply watch all newsgroups for a value with the right
format, much as Mr. Kibo watches for his name in postings everywhere.
Knowing that the delivery is taking place, it can be stopped, but it's still
going to be very hard to trace it to some perpetrator.  However, the
perpetrator still has to have a way to USE the credit-card numbers thus
obtained, and that's the point at which traditional fraud detection and
law-enforcement measures become relevant.

Re: RISKS (...) of typing credit-card numbers (Sibert, RISKS-17.69)

Fisher Mark <>
Mon, 12 Feb 96 08:15:00 PST
Although Mr. Sibert understands the overall context of Nathaniel
Borenstein's original message, there are a few points on which I would

1. Nathaniel's original message was going to state, "NEVER TYPE YOUR CREDIT
CARD NUMBER INTO AN INSECURE COMPUTER".  Because computer security is hard
to understand for those who feel good just being sure which icons are the
minimize and maximize boxes (i.e. most computer users), he dropped the
"INSECURE" from his message.  Unfortunately, most computers now in use
(which means PCs) run operating systems that are insecure (MacOS, DOS,
Win3.1, Win95).  Although Windows NT is presently vulnerable to this attack,
a simple additional API call that retrieves a string from the user that is
not visible to other programs or hooks would prevent the attack.  NT does
this kind of operation now for logins.

2. Although most virus software is primitive ("Sturgeon's Law: 90% of
everything is junk"), there is no reason for this situation to continue when
such an immense financial gain can be made.  How many of us _really_ know
all what software we are running at any given moment?  Win3.1, DOS, and I
believe Win95 do not normally come with a program that can list all running
programs — and I suspect the same is true of MacOS.

3. Communicating the credit-card data to another system could be done via
Windows Sockets TCP/IP API calls, which (if there is already an active
connection) could be done entirely silently in the face of all errors.
 Since "netstat"-like utilities can list all active TCP/IP connections,
determining whether or not there is already at least one connection ready to
be piggybacked is a solvable problem.  Even with the inevitable
compatibility problems, a lot of credit cards could be stolen in just a
week's time.

4. Widespread distribution could likely be achieved by suitable social
engineering — adding the Trojan Horse to one of the interminable series of
CompuServe CD-ROMs/diskettes I keep receiving, for example.  I would suspect
that most people who actually package software don't go through the kind of
background checks I did to get my security clearance...

The real solutions to these combinations of attacks are:
1. Run a secure OS; and/or
2. Deal with vendors that do not use a self-verifying ID for purchases; and/or
3. Use secure hardware to transmit/generate the user's purchase ID.

Mark Leighton Fisher, Thomson Consumer Electronics, Indianapolis, IN

Leahy to introduce bill to repeal CDA!

Stanton McCandlish <>
Thu, 8 Feb 1996 21:07:17 -0800 (PST)
[From (D-VT) Sen. Pat Leahy's WWW Pages, specifically at:
It is advisable to send e-mail, and make phone calls, to his office in
support of his upcoming legislation to repeal the Communications Decency
Amendment to the Telecom Bill.]

              U.S. Senator Patrick Leahy - why.html

 I am turning my World Wide Web page black for 48 hours as part of the
 protest by my fellow Internet users against the online censorship
 provisions of the new telecommunications law. The online censorship
 provisions should be repealed, and I plan to introduce legislation to do
 just that.

 I was one of five Senators who voted against this legislation, in large
 part because of what I believe are unconstitutional restrictions on what
 we can say online. I hope you will take the time to read my full
 statement on the telecommunications law contained on my web-page. While I
 do not condone the transmission of obscene material or child pornography
 on the Net, I believe the solution proposed in the telecommunications law
 will do more to harm the use and growth of the Net without combatting the
 problem for which it is intended.

Sen. Patrick Leahy (D-VT)

Foreign `replies' cause anxiety

Timothy Mowchanuk <>
Sat, 10 Feb 1996 18:56:10 +0000
A bit of background first.  There is a disorder called `high functioning
autism'.  These people are usually quite intelligent, (one is even an
Associate Professor) but have have trouble with social and emotional
relationships.  They tend to interpret events and personal reactions poorly
and very literally.  They can be very sensitive about this deficiency.

While monitoring an Autism list an interesting event occurred.  There was an
ongoing argument (not a flame war) between one member (self-confessed high
functioning autistic) and the rest of the group.  Suddenly she sent a
message to the group that he/she was unsubscribing.  Apparently there had
been some strong personal e-mail to this person including *several messages
in a foreign language*.  This person was *quite* upset that someone was
apparently sending `flames' in a foreign language.

After several days the list finally figured out what was happening.  There
was a subscriber in Brazil who had not check his/her mailbox and it was
full.  It turns out that the Brazilian majordomo was sending return messages
to the effect that this mailbox was full.  (I don't understand why this
person, and only a few others, got the messages and not the sending
listserver so don't ask.)

This person is/was very sensitive to e-mail messages.  In this case messages
in a foreign language cause a significant amount of emotional distress.
There are a significant number of electronic emotional support groups.  By
definition, these lists will contain a large percentage of people with
emotional/mental problems. Is this a risk?  If so, what is the solution?

Timothy Mowchanuk, Technology Education, Queensland University of Technology
Brisbane, Queensland, Australia

Correction: Train operators get permission to use manual backup (17.70)

Jonathan Kamens <>
Fri, 9 Feb 1996 14:05:23 GMT
   From: "Tom Comeau @ Space Telescope Science Institute" <>
   Supervisors must approve taking a train out of manual mode,
I believe this should read "automatic mode".
Jonathan Kamens  OpenVision Technologies, Inc.

   [Darn, and I thought I had proofread very carefully.  You're correct.
   (Though actually, supervisors until last Sunday had to approve both
   going into and out of manual mode, so I guess we're both right ;-)  )
   Tom Comeau]

Re: Electronic Medical Records and Images (Brown, RISKS-17.70)

David Coburn <>
Mon, 12 Feb 96 14:29:48 PST
Jay Brown points out that

> One potential risk with this type of system would be the "configuration
> management" challenge - what image goes with what patient?

I would point out that this isn't too much of a risk.  This is a classic
database problem, and providers of these services are for the most part
using commercial RDBMS solutions, including those of my employer, to
overcome these sorts of risks.

I would have actually seen this as _reducing_ the risk of misidentification
of the information, as opposed to an increase.  It is rather hard to drop a
video tape or CD ROM and have the pictures spill out on the floor, after
all.  Having worked many years in a hospital, I can ASSURE you that this is
not currently the case.

David Coburn, Informix Software, Inc. ...uunet!infmx!coburn

Re: Electronic Medical Records and Images (Brown, RISKS-17.70)

Allan Noordvyk <>
Fri, 9 Feb 96 08:06:06 -0800
As a software developer at a company which makes filmless medical imaging
systems, I can say that this risk is well understood and has been dealt with
by a variety of vendors for quite a number of years.  In fact, the news item
which prompted Mr. Brown's article describes deployment of a system which is
not terribly new or revolutionary with the exception that it deals with the
bandwidth-hungry full-motion cardiac video.  Our company, for example, has
customers have been operating completely filmless for over four years.

In addition to strong database integrity measures and a robust
fault-tolerant message model, we, like most PACS (Picture Archiving and
Communications Systems) vendors, take great pains to embed patient and exam
identifying information into each series of images in *addition* to the
information stored in the indexing database.  In the event of some sort of
horrible disaster, a known good back-up the database can be loaded in place
of a damaged one and then the more recent information determined by a walk
of the on-line and (if necessary) off-line archives.

Furthermore, the new DICOM (DIgital Communications in Medicine) standard
which is used by different vendors to network their systems in fact this
makes required in the header for each image itself.  The overhead required
to do this is minimal relative to the size of each medical image (range 256
x 256 8-bit grayscale, through 640 x 484 24-bit color, to 2K x 2K 16-bit
grayscale) and the fact that lossy image compression algorithms generally
can't be used if the image is to remain diagnostic quality.

The risk of the wrong images being associated with a patient actually
predates computerization of radiology.  Given the sheer number of sheets of
film generated each day at a radiology department and the resulting
likelihood of the sheets going astray, most modalities long ago adopted the
practice of burning the name and identification of the patient *visibly*
into each image itself.  I expect this practice to continue in the digital
age.  Thus there is an immediately visible human check on the operation of
any computerized system in addition to "Why does Mr. Johnson have a uterus?".

All in all, I would say that there is more of a risk of being wheeled into
the wrong operating room than having a radiologist accidentally diagnose
your case from someone else's images.

Allan Noordvyk, Software Artisan, ALI Technologies  Richmond, Canada
Voice: 604.279.5422 x 317    Fax: 604.279.5468

  [Postscript added later by Allan:]

While the original proposed risk (i.e., the wrong images being associated
with a patient) is quite low (as discussed in my previous missive), there is
another slightly different risk which is more likely and has just a large
potential for harmful results to the patient.  The risk is that when a
radiologist is looking at digitally transferred images of a patient's exam,
can he or she be absolutely sure that *all* of the images captured are
actually being shown?  A fairly serious medical condition can often be
evidenced in only a single image from a large set, and thus omission of one
or more images runs the risk of a misdiagnosis (eg. a false negative).

The basic network dependability problem is complicated by the possible
presence of WANs (eg. we have a number of sites where outlying clinics feed
images to expert radiologists at a central institution).  These
lower-bandwidth conduits can introduce significant lags in the transfer of
large images.  Unexpected communications difficulties (eg. momentary Telco
line outage) may occasionally increase these lags without the knowledge of
the users of the system.

My company's system contains extensive interprocess status messages and
sanity checks to ensure that this risk is virtually eliminated.  However
this only works when our system handles both the capturing and viewing of
the images at either end of the network as the protocols are a careful
proprietary design developed over the years.  With the wide-spread adoption
of DICOM as the lingua franca by which different vender's medical imaging
equipment uses to talk amongst themselves we are seeing heterogeneous
deployment of components in imaging networks.  Unfortunately the DICOM
standard has, up to now, avoided a lot of the hard issues involved in
handling this risk.  Recent proposed additions to the standard address the
problem somewhat, but these would be optional for vendors to implement.
Thus there is always the risk of a minimalist DICOM implementation breaking
end-to-end robustness of more complete systems with which they are

Re: Electronic Medical Records and Images (Brown, RISKS-17.70)

Tom Olin <>
Sun, 11 Feb 1996 09:31:59 +0500
I suppose it is worth repeating the obvious risks of life from time to time,
but after a while, that approach turns RISKS into a pretty monotonous,
boring, and ultimately worthless publication.

The precise same risk for electronically stored and transmitted medical
images could just as easily be applied to any type of information storage on
any type of media.  Consider a slight rewording of the last sentence quoted
        As any secretary knows, keeping track of a lot of different
        files in a filing cabinet, many of which have more than one
        version, is not a trivial task.

Or take it back to the first cave dweller who tried to keep track of
his kids on stone tablets:

        As any caveman knows, keeping track of a lot of different kids
        on stone tablets, many of which have more than one version, is
        not a trivial task.

Too often, I think one of the risks of RISKS is that too many people try too
hard to find risks in every little thing.  Perhaps you (the moderator) and I
disagree on what risks are worth mentioning, but it seems that you usually
keep things on the interesting side.  Not this time.

Or maybe I'm just grumpy because I've been working too many weekends lately.

Tom Olin, PAR Government Systems Corporation, 8383 Seneca Turnpike,
New Hartford, NY  13413-4991  +1 315 738 0600 x638

Please report problems with the web pages to the maintainer