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 5 Issue 70

Sunday, 6 December 1987

Contents

o Wall Street crash, computers, and SDI
Rodney Hoffman
o NW Flight 255 -- Simulator did, but wasn't
Scot E. Wilcoxon
o Whistle-blowers who aren't
Henry Spencer
o Re: Space Shuttle Whistle-Blowers Sound Alarm Again
Henry Spencer
o A new twist to password insecurity
Roy Smith
o More on PIN encoding
Chris Maltby
o Telephone overload
Stephen Grove
o Software licensing problems
Geof Cooper
o Re: Mariner 1 or Apollo 11?
Henry Spencer
Brent Chapman
o More on addressable converter box
Allan Pratt
o Centralized car locks
K. Richard Magill
o Info on RISKS (comp.risks)

Wall Street crash, computers, and SDI

Rodney Hoffman <Hoffman.es@Xerox.COM>
6 Dec 87 12:29:47 PST (Sunday)
From the 'Letters' column in the 'Wall Street Journal', Thursday, Dec. 3, 1987:

        SDI COULD BE TOO SWIFT

The role of computers in the recent wide fluctuation of stock prices brings to
focus an interesting issue that has wider implications.  The initial downward
trend in the market was greatly amplified by rapid computer-initiated program
trading.  From our long experience in computer science and dynamical systems
analysis, we fear that this dangerous amplification effect could occur in other
more critical computational networks, most notably those envisioned for the
strategic defense initiative (SDI).

SDI planners have proposed giving computers a key role in the decision about
retaliatory measures in the event of attack.  Others have argued persuasively
that a computer program that reflects our policy cannot be reliably constructed
or completely debugged.  But setting this aside, we claim that there would be
great danger in the very speed of such programs, unmodulated by slower hauman
interactions that provide effective damping.  It is known that nonlinear
systems (such as the ones composing computer networks) can amplify very small
disturbances.  As illustrated by the program trading example cited, this can
cause massive changes in overall system behavior unplanned for by system
designers.

The programs in the defense network for SDI must process incoming signals for
possible threats, and act rapidly in accord with the resulting analysis.  Any
overt action by the SDI system can lead to a rise in the readiness of the
opposite side; blasts in space can be interpreted by Russian programs as
attacks on spy satellites, a preparatory move for a U.S. first strike.  This
automatic feedback loop, through the Russian and American computers and
sensors, can easily amplify the intesity of a dangerous situation to the point
of nuclear catastrophe.  In order to dampen such an inadvertent escalation,
humans must be involved in the response progress, even at intermediate stages.
They must have time enough to think and communicate to avoid the nonlinear
amplification effects.  In the case of the envisioned system for SDI, we
believe that there is no effective way whereby people can modulate the behavior
of a computer system while retaining the hoped for rapid response.

                        Daniel G. Bobrow
                        Bernardo A. Huberman
Palo Alto, Calif.


(NW Flight 255) Simulator did, but wasn't.

Scot E. Wilcoxon <umn-cs!datapg.MN.ORG!sewilco@cs-gw.D.UMN.EDU>
Thu, 3 Dec 87 13:36:36 CST
The random failure of a $13 circuit breaker may have contributed to an
airplane crash.  Also, an indicator in the simulator for the aircraft
behaves differently than in the real aircraft when that failure occurs.

Northwest Flight 255 apparently crashed in Detroit three months ago because
the flaps were not lowered during takeoff.  The MD 80's takeoff warning
system should have warned the pilot, but its audio warning "flaps, flaps"
was not on the cockpit recording so the warning system probably failed.

NTSB examination of a $13 circuit breaker that supplies power to the
warning system has found several planes with breakers that would not
pass the electrical current that they should.

A McDonnell Douglas document states that when power fails to the warning
system, a warning light (CAWS fail light) should go on in the cockpit.
That is what happens in the MD80 simulator, but in an actual MD80 aircraft
the warning light does not go on.

A McDonnell Douglas official said the document is "clearly in error".  (quoting
StarTribune:) "This revelation has put the FAA in the awkward position of
ordering changes that will make the simulators behave the same way as the
airplane instead of making the airplane behave like the simulator."

The simulator will be altered instead of airplanes because the warning
system is classified by the FAA as not requiring backup systems.  The
warning system is required (FAA won't allow takeoff if it is not working),
but is considered "nonessential" for manufacturing purposes (FAA does
not require backup systems for nonessential systems).

(Information from 11/28/87 Minneapolis Star Tribune, pg 1,4D)

Scot E. Wilcoxon    sewilco@DataPg.MN.ORG   {ems,meccts}!datapg!sewilco
Data Progress       Minneapolis, MN, USA    +1 612-825-2607


Whistle-blowers who aren't

<mnetor!utzoo!henry@uunet.UU.NET>
Wed, 2 Dec 87 15:47:52 EST
>  Maxson will share the stage with former Morton Thiokol engineer Roger
>  Boisjoly, who currently has a billion-dollar suit underway...

Maybe I am just being picky about this, but it still makes me see red when
I see Boisjoly described as a "whistle-blower".  Boisjoly is the man who
could have blown the whistle BUT DIDN'T, and seven astronauts died as a
result.  Boisjoly was the engineer who told MT management "don't launch",
was told "put on your management hat", did so, and changed his expert
professional opinion 180 degrees to match his hat color.  In a just world,
I cannot help but think that he (and, certainly, his management) would be
facing criminal charges.  Boisjoly did not blow the whistle; he merely
turned "state's evidence" after the fact.

Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,decvax,pyramid}!utzoo!henry


Re: Space Shuttle Whistle-Blowers Sound Alarm Again (reprint)

<mnetor!utzoo!henry@uunet.UU.NET>
Wed, 2 Dec 87 15:48:00 EST
> ... new and improved shuttle escape mechanisms. Lot's of
> money is being spent, but whether reported or not, upon (close) examination
> none of these mechanisms would prevent the death of astronauts in a
> Challenger type disaster.  I wonder just how much additional engineering
> is happening for purely public relations purposes...

The escape work is not being done for purely public relations purposes; it
merely, for the most part, does not address situations as severe as the
Challenger disaster.  There is in fact some attention being given to such
situations, but the thorough re-examination of shuttle safety issues turned
up other cases where modest effort would yield a much higher probability
of survival.  The reason why most escape-system work is not addressing the
Challenger scenario is that it is very difficult to get the crew out
of such a situation reliably!  There are also tradeoffs to be considered:
regardless of managerial idiots blithering about safety being an absolute
priority, the only way to make the shuttles completely safe is to put them
in museums and never fly them again.  In practice, there is no way to avoid
some level of compromise between safety and utility, since adding any type
of escape system reduces payload.  There are also safety-vs-safety tradeoffs
to be made, since even simple ejection seats can and do fire accidentally,
often with fatal consequences.

Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,decvax,pyramid}!utzoo!henry


A new twist to password insecurity (human factors)

Roy Smith <roy%phri@uunet.UU.NET>
6 Dec 87 00:19:06 GMT
A bunch of people around here have signed up for the BRS/Colleague
bibliographic data base service.  Each person has an individual account
number and (supposedly secret) password.  We get a combined invoice each
month, with usage itemized and identified, but only by account number (not
by name).  Our accounting office didn't know what to do with the account
numbers, so I called BRS and asked for a list of which name corresponds to
which account number.  Much to my surprise, I got in the mail a few days
later a list of names, account numbers, *and passwords*.

Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016


More on PIN encoding

Chris Maltby <munnari!softway.oz.au!chris@uunet.UU.NET>
2 Dec 87 11:47:01 +1100 (Wed)
A recent fraud case in Sydney reveals that the PIN details are either
encoded on the card, or a function of the card number. The perpetrators of
the fraud used some ingenuity in their system.

The first stage was to deduce the magnetic encoding on the card's strip from
discarded receipts collected around ATMs, and then manufacture cards which
duplicated the real card. Unfortunately for them, they were unable to break
the PIN encoding algorithm, so they resorted to hanging around in their cars
opposite the ATM location with a portable video camera and a zoom lens. When
some unsuspecting user left behind a recipt card they were able to
re-manufacture his card and replay the video of his PIN entry.

There are several morals. First - don't leave your receipts at the machine.
Second - stand close when entering your PIN. Third - don't use the system at
all - can you trust any system which is this easy to break. If the forgers
had been just a bit more resourceful they would have decoded the PIN as
well.  Of course, the "conditions of use" of your card make you liable for
such frauds...

Chris Maltby - Softway Pty Ltd  (chris@softway.oz)

PHONE:  +61-2-698-2322   uunet!softway.oz!chris   chris@softway.oz.au


Telephone overload (Re: RISKS-5.63)

Stephen Grove <ptsfa!pbhya!seg@ames.arpa>
Wed, 2 Dec 87 16:09:35 PST
> Date:         Fri, 20 Nov 87 15:17:09 PST
> From: "LT Scott A. Norton, USN" <4526P%NAVPGS.BITNET@wiscvm.wisc.edu>
> Subject:      L.A. Earthquake & Telephone Service
> 
> Can anyone with better knowledge of the phone companies' local offices tell
> me if there is some simple way to shed this extra load in a reasonable way?
> I know that after some minutes off the hook, the phone loses its dial tone.
> Does this adequately release the resources the off-the-hook phone was using?

The older electromechanical systems, required someone to throw a switch and
remove the battery supply from the non-priority customers.
Priority-customers being those in the class of Hospitals, Police, Fire, etc.

The newer ESS offices (ESS = Electronic Switching Systems, using stored
programs for control, as opposed to hard-wired logic) determine when the
load is getting excessive, and delay the response to nonpriority customers
by a factor of three (I think). In other words if the ESS normally responds in
300ms, it will now take 900ms. I have seen it work in a flood, and it worked
fine.  It was inhibited at first, and the ESS repeatedly stated the need to
implement the control, and was unable to serve anyone, but when allowed,
response was slower, but the calls went through.

For call in programs and promotions, we like to provide special prefixes
that can be limited in the number of interoffice channels they can access.

     Stephen Grove, Pac Bell, Rohnert Park, Calif.
        UUCP:{ihnp4,dual,sun,hoptoad}!ptsfa!pbhya!seg


Software licensing problems

Geof Cooper <imagen!geof@decwrl.dec.com>
Wed, 2 Dec 87 12:47:00 pst
We too have experienced the problem of software licensed to run on only some
nodes.  Our approach has been to convince the repairman types to switch the
Node-ID rom's on the apollos in question, so that your access key follows
you where you go.  I've never seen the address ROM itself fail!

Apollo Computer has recently brought out a product that allows N copies of a
program to be running simultaneously on any of some larger number of
machines.  It remains to be seen if any vendors will be interested in using
the product.  From the vendors' point of view, there is a possible financial
risk to adopting the new scheme, since many programs are used less than full
time; the customer might elect to buy fewer copies of the program and take
the risk that occasionally someone will have to "wait for a dialtone".
E.g., we have about 25-30 licenses for Interleaf's desktop publishing
software.  I've never seen more than 15 of them in use at once).
                                                                  - Geof


Re: Mariner 1 or Apollo 11? (RISKS-5.63)

<mnetor!utzoo!henry@uunet.UU.NET>
Wed, 2 Dec 87 15:48:18 EST
>     I heard that the famous "./," disaster caused the problem with the
> onboard IBM 1800 on Apollo 11...

The onboard computers on Apollo were not IBM 1800s unless I have confused
the numbers badly, and almost certainly they were programmed in assembler
due to severely limited ROM capacity, so I'd be a bit skeptical of this.

                Henry Spencer @ U of Toronto Zoology
                {allegra,ihnp4,decvax,pyramid}!utzoo!henry


Re: Mariner 1 or Apollo 11?

Brent Chapman <chapman%mica.Berkeley.EDU@violet.berkeley.edu>
Mon, 30 Nov 87 22:26:10 PST
In RISKS-5.64, Scott Dorsey <kludge@pyr.gatech.edu> writes:

>    I heard that the famous "./," disaster caused the problem with the
>onboard IBM 1800 on Apollo 11.  I heard this from a professor who teaches
>Fortran, so I'm not so sure about the reliability of the source.  Anyone
>else have information on either the Apollo or the Mariner problems?

If you mean the "decent alarm" that occurred in the Lunar Module moments
before lunar touchdown, then one of the NASA documentaries (unfortunately,
I don't remember the title, or even when or where I saw it) told a different
story.  They said that the alarm was caused by an overload condition in the
processor; apparently the Armstrong and Aldrin had left a certain sensor
(radar altimiter, I think) enabled that was supposed to have been shut down
by that point in the landing sequence, and the added load of that sensor
caused the computer to falsely register an alarm condition.  If I remember
correctly, the programmer that had written the piece of code involved was
in Mission Control at the time the alarm occurred; he stared at the situation
and status boards for a few seconds, then announced that he knew what the 
problem was, that it was a false reading, and advised to continue the landing
(although I'm not sure if they could have aborted at that stage or not).

I may be confusing what I saw in the documentary (that the false decent alarm
happened because the processor was overloaded because a sensor that should have
been turned off wasn't) with "urban legend" (that the programmer responsible
was in Mission Control, etc.); can anyone else back me up on this?

Brent Chapman                   Capital Market Technology, Inc.
Senior Programmer/Analyst           1995 University Ave., Suite 390
{lll-tis,ucbvax!cogsci}!capmkt!brent        Berkeley, CA  94704
capmkt!brent@{lll-tis.arpa,cogsci.berkeley.edu} Phone: 415/540-6400

   [I am sorry that misinformation is still flowing on this subject.  I have
   held up a large number of potential contributions to RISKS, awaiting a
   definitive report that is rumored to be working its way RISKSward.  PGN]


More on addressable converter box

Allan Pratt <ucbcad!ames.UUCP!atari!apratt@ucbvax.Berkeley.EDU>
Mon, 30 Nov 87 11:40:31 pst
RISKS@KL.SRI.COM (RISKS FORUM, Peter G. Neumann -- Coordinator):

I have a note to add about my addressable converter box: the hardware is
two-way capable.  I know this because there is an "event" button which
you hit to purchase an event on Pay-Per-View.  You then key in your
purchase-authorization code, and you get the event.  Obviously, the box
has to be two-way, because the cable company has to be able to bill you. 
Now, the particular cable company in my area does not support this, but
the box hardware & software do.  The manual talks about it, with an "If
your cable company supports it" disclaimer. 

Opinions expressed above do not necessarily -- Allan Pratt, Atari Corp.
reflect those of Atari Corp. or anyone else.      ...ames!atari!apratt


centralized car locks (foaf)

K. Richard Magill <umix!oxtrap!rich@RUTGERS.EDU>
30 Nov 87 18:17:47 GMT
It's my understanding that a certain lesser known car (the Bricklin) was
sold with entirely electronic locks.  When the battery died or shorted you
were entirely locked out.  The Bricklin had gull wing doors.
                                                                rich.

                                 [When the gulls attacked, the car owners 
                                 were known as the Bricklin drudgers.  PGN]

Please report problems with the web pages to the maintainer

Top