The RISKS Digest
Volume 16 Issue 39

Tuesday, 6th September 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…


PKZIP encryption broken (known plaintext attack)
Paul Carl Kocher
Some privacy notes
Phil Agre
Database Marketing (privacy in *Business Week*)
Mark Stalzer
Backspace Problems
John Vilkaitis
Backspace Failure
John Vilkaitis
Re: Millenium goes to prison
Jim Hiller
_Modern_ risks of call by reference
Mike Albaugh
Some comments on the A330 accident
Peter Ladkin
ESORICS 94 Program
Yves Deswarte
Info on RISKS (comp.risks)

PKZIP encryption broken (known plaintext attack)

Paul Carl Kocher <kocherp@leland.Stanford.EDU>
Sun, 4 Sep 1994 17:31:28 -0700
I finally found time to take a closer look at the encryption algorithm
by Roger Schlafly that is used in PKZIP and have developed a practical
known plaintext attack that can find the entire 96-bit internal state.

The basic encryption algorithm has four steps, two of which are based on
linear shift registers, one is like a linear congruential, and the final
converts the contents of an internal state register into an 8-bit value to XOR
onto a plaintext byte.  A complete description of the algorithm is included in
the file APPNOTE.TXT, which is included with PKZIP version 1.1 (check Archie
for "pkz110.exe").

Although the algorithm is substantially better than the toy ciphers used in
many products, I have developed a practical known plaintext attack that finds
the 96 bit internal state.  Unlike the ZipCrack program I released a couple
years ago, this attack finds the internal state registers directly and does
not involve a brute-force attack on the password.  If adequate known plaintext
is available, my attack will find the state, regardless of the password's size
or content.

My attack is an improvement on a known plaintext attack described in a paper
by Biham (unpublished work) that takes 2^38+ operations.  My improvements
reduce the amount of work required by approximately a factor of 1500 with 200
bytes of plaintext.  With less plaintext the attack will take somewhat more
time, but just 40 bytes should be enough to be practical.  I've written code
for all steps of the attack; a version written in C with a few optimizations
in inline assembly runs in less than a day on my '486.  The attack will work
with versions 1.1 or 2.xx of PKZIP and other programs using the same

A more in-depth description of the attack will be made available soon, but I
wanted to let people using PKZIP (and any other programs that use the same
algorithm) know immediately about the weakness.

Paul C. Kocher  Independent data security
  consultant/contractor.  415-323-7634  [Disclaimers removed.  PGN]

Some privacy notes

Phil Agre <>
Mon, 5 Sep 1994 18:37:31 -0700
The September issue of *Smithsonian* magazine includes a long article on
"ubiquitous computing" research at Xerox, with some attention to the moral
issues relating to tracking and monitoring.

The 5 Sep 1994 issue of *Business Week* has a cover story on database
marketing.  Like most *Business Week* cover stories it's a superficial rehash
of items you might have seen elsewhere.  But it might be useful as a summary.

Finally, here is a wonderful quotation from a much longer article by Edwin
McDowell, ``The scrambling is on for off-season tourism'' (*The New York
Times*, 5 Sep 1994, business section, pp. 17-18) on off-season tourism

  "Another reason for the growing success of off-season strategies is that
  "states have become a lot more sophisticated with their data bases", said
  James V. Cammisa Jr., a travel industry consultant in Miami.  "They know
  where the peaks and valleys in their tourism operations are, and they know
  how to market the off-season effectively.

  "Kentucky's data base showed that only 350,000 of the 2.5 million Canadians
  who drove through the state last year stayed overnight.

  "Our research showed that 83 percent of them come from January to
  June, headed for Florida, South Carolina and the beaches of Alabama and
  Mississippi", said Robert Stewart, the Commissioner of Travel Development
  for Kentucky.  To entice more of them, Kentucky officials will soon hold
  a press conference in Toronto and Canadians will be offered a card giving
  them discounts at hotels, restaurants and attractions along three of
  Kentucky's interstate highways.

  "Also for the first time, Kentucky is using direct mail to bolster anemic
  winter occupancy rates in its 15 resort parks that offer overnight
  accommodations year-round."  (page 18)

This kind of database marketing is worth thinking about in the context of
rapidly advancing proposals for thoroughgoing instrumentation of cars and
roads under the rubric of "intelligent vehicle-highway systems", particularly
given that most of the marketing organizations mentioned in the article are in
fact government agencies using commercial methods for the benefit of private

Phil Agre, UCSD

Database Marketing

Tue, 6 Sep 1994 13:44:22 +0800
The cover story of the current issue of Business Week (5 Sep 1994), a
conservative business magazine (sorry, Phil), is on Database Marketing.  The
goal of Database Marketing is to build detailed customer profiles so that a
company can target advertisements to specific customers for products and
services. This approach is highly successful: response rates are double digit
as opposed to 2%--3% for junk mail.

The data collection process starts with a customer's past purchases.  Other
sources include surveys, rebate requests, and warranty cards.  American
Express scans a customer's individual transactions to find patterns and to
suggest local places that take the card.  Many hospitals sell the names and
addresses of families with newborns.  The data is then combined with public
records, such as drivers' licenses, auto registrations, and property tax
rolls.  Ohio sold its drivers' license and car registration lists for $375,000
to TRW.  What results is a detailed profile of each customer.

The computing technology used to mine a database for prospects includes
parallel processing and neural networks. Neural nets are trained to look for
people likely to buy a product or service given the parameters in the
database, e.g.,

  what combination of income level, investment activity, and credit-card
  spending is most likely to be seen among people who are in the market for

The net is applied against each profile in a process called "drilling down."
This is a compute intensive operation and companies are starting to resort to
parallel processing or workstation clusters.  Indeed, it's estimated that a
large portion of the projected growth in commercial parallel processing, from
$400M today to $5B in 98, will be for database marketing applications.

When asked about the privacy issues, one marketer responded that the loss of
privacy is offset by the convenience to the customer of highly selective
advertising. I'll forgo the commentary and simply refer the interested reader
to the original source for more details and anecdotes.

Mark Stalzer,

Backspace Problems

Javilk <>
Sat, 3 Sep 1994 04:28:10 -0700 (PDT)
I subscribe to the NETCOM Internet service provider.  Although new to UNIX, I
have been using computers for over 25 years, and had once worked for one of
the largest timesharing service providers in the world.  I currently work as a
consultant in the areas of software development on PC's and mainframes.

On a number of occasions while writing E-mail using NETCOM's MAIL utility, I
was chagrined to discover that my backspace/DEL key has been disabled,
yielding a "^?" rather than deleting the previous character.  This problem
also occurs at the UNIX command level, yet utilities, such as the PICO editor
and slower to use ELM E-mail utility interpret the backspace/DEL key
correctly. (Except for the subject line!)

Further investigation suggests the once this failure occurs on a server, it
remains till corrected by the staff.  Reloading my telecom package changes
nothing, getting another server does; but there is no means of selecting which
server will answer my call.  My E-mail complaints to NETCOM yielded various
responses from a rather insistent and unfriendly person regarding a "^H" in
the text, and flaming me as incompetent when I tried to point out that I
neither typed a "^H" in the E-mail, nor see it on my screen, and do not find
it consistent across servers; to several more gentle statements by other
persons to the effect that there are some differences between servers but that
is "Not a Problem", "not a bug" — if you see a "^H", go fix your terminal
program.  I have even had one E-mail me that if I was not satisfied, I should
change to another internet service company.

     (For the last time, I see a "^?", NOT a "^H"!  And not on every
server!  If it is MY problem, how is it that their staff fixes it on
their end every once in a while — even when I don't complain?)

The RISK of not reading E-mailed complaints is finding it in RISKS.

The RISK at command level is entering unintended ambiguities when attempting
to correct parameters.

The RISK is sending out correspondence with garbage characters and unmeant
words.  (As in earlier correspondence with the Moderator.)

And as a person who started in the field on a help desk in 1970, the GAIN in
LISTENING to what customers say, ESPECIALY when everyone else is telling them
to go away, is making good friends.  Their problems usualy turn out to be Very

-JVV- (  John V. Vilkaitis, Senior Consultant
Software General Corp.   Field Office: 408-983-0518 (Voice/Fax)

Backspace Failure

Javilk <>
Sat, 3 Sep 1994 04:53:00 -0700 (PDT)
This problem reminds me of another problem I had as a student working on the
old IBM 360, where I would occasionaly see this error on my ONLINE-OS 2250(?)
video terminal:

   (IBM-ese number) ILLEGAL ERROR.

whereupon my partitioned data set would be trashed.  There would be NO
hardcopy of the message.  The center staff would tell me, with varying degrees
of politeness, that I was "working too hard" or "staying up too late".

   Finally, after months of occasional problems, I spent one night looking
through A LOT of manuals to find the explanation that the error routine had
found an illegal pointer in the traceback chain, and thus it was the error
information that was "illegal".

   The first step in solving a problem, is listening to the person having
the problem.   How can you solve a problem, when you don't know what it is?

-JVV-  John V. Vilkaitis, Senior Consultant
Software General Corp.  Field Office: 408-983-0518 (Voice/FAX)

Re: Millenium goes to prison (RISKS-16.37)

Tue, 6 Sep 1994 21:05:48 -0400 (EDT)
I have to wonder whether there was any intended marketing connection
between the name chosen for the above-referenced communication system
(Millenium Inmate System) and the resulting acronym...

One can derive a lot of humor envisioning a Millenium press release
extolling the virtues of the system, but using only the acronym, and
then applying the discussion out of context in a community of automators!

The RISK here isn't entirely intuitive, but smells something like the
risk of choosing a product name without regard for semantics that might
be invoked by a segment of the product's potential market...

Jim Hiller

   [Something is aMISs?  In this case, a MIS is as good as a smile
   (from a .MIL source, at that!).  PGN]

_Modern_ risks of call by reference

Mike Albaugh <>
Tue, 6 Sep 1994 10:16:53 -0700 (pdt)
    I realize you are probably sick to death of the "3=4" thread,
but the thing that struck me was that all the contributions were of
the "When I was a kid we had to walk ten miles through the snow and
use a compiler that could bung up its constants" form. What saddens
me is that the introduction of the "reference" operator in C++ indicates
that computer science has apparently _not_ learned the lesson taught
by the earlier and very well documented problems. It is simply not
advisable to have a single character buried in a 1000+ line "include"
file radically change the behavior of:

    double my_angle,result1,result2;

    /* we can't make my_angle const, because it needs to be
     * "tweaked" on a per-run basis, so neither prototypes
     * nor MMU's can save us...
    my_angle = get_current_operating_assumptions();

    result1 = some_library_function(1,my_angle);
    result2 = some_library_function(2,my_angle);

    In C, one can be confident that no matter what else mat be
wrong with some_library_function(), it will _NOT_ damage my_angle.
In C++, the addition of a single '&' character destroys the basis
of that confidence. I can forgive Backus for "changeable constants",
but Stroustrup should have known better :-)

    The average sailor will not spit into the wind a second time. The
average computer scientist does not, apparently, learn from experience.

Mike Albaugh, Atari Games Corp (Arcade Games, soon Time Warner Interactive)
675 Sycamore Dr. Milpitas, CA 95035   (408)434-1709

Some comments on the A330 accident

Peter Ladkin <>
Sat, 27 Aug 1994 19:02:00 +0200
There are a few points worth emphasising which follow from the Air et Cosmos
issue 1482 summary of the A330 accident preliminary report, along with the
1480/1 AeC summary of the preliminary-preliminary findings from the telemetry

The A330 preliminary accident report singles out lack of pitch
protection with the autopilot in ALT* mode as a determining factor.

According to the report by Casamayou in Air et Cosmos 1480 (11-16 July), the
copilot rotated to 28deg to hold 150kts of speed (the airplane actually went
to 29deg), and the autopilot was engaged by Warner, who also retarded the left
engine and cut the left hydraulic pump to simulate an engine failure: `As
planned, the pitch of the aircraft started to diminish and passed from 29deg
to 25deg, the [pitch] limit authorised by the [flight] envelope protection
system FMGES (flight management guidance and envelope system).'

It is presumed that the pilots were expecting that the autopilot was to remain
in SRS mode (`Speed Reference System') under which there is automatic pitch
protection.  However, because the altitude was set too low (2000ft) in the
flight director (FCU), the autopilot reverted almost immediately to ALT* mode,
under which there is no pitch protection.  However, it was non-obvious for the
pilots to know they were in ALT* mode since it wasn't displayed on the PFD
under those flight conditions - mode info disappears from the PFD at 25deg,
**the same point to which pitch is protected by the FMGES**.

The preliminary report noted the lack of PFD display of mode as a contributing
factor, but not a cause.  Bernard Ziegler, technical director of Airbus,
singled out in interviews the action of achieving 25deg of pitch as one of his
main contributing factors [RISKS-16.35, also the specific figure of 25deg, a
`particularly high pitch angle' is found in Flight International, 17-23 Aug
1994, p4]. (The other two factors mentioned in the Speigel interview were the
2000ft altitude setting and that the pilots waited too long to recover.)

However, if you want to test pitch protection it follows you have to put the
airplane into more than 25deg of pitch, which is what the pilots did.  But
this is a flight condition such that you can't tell on the PFD what AP mode
you're in, and hence whether pitch is actually protected!  This info might be
available, but it is not displayed on the PFD.

Contributory factors that were also noted by the report: the full-aft center
of gravity, and the TOGA thrust on the engines. However, the airplane may be
legally loaded to full-aft CG, and if a go-around is needed on an automatic
landing, that's what TOGA thrust is for. TOGA conditions are statistically the
most likely conditions under which there is an engine failure.

All of the above is a matter of record, or of common knowledge.  I'd like to
add a few comments and questions of my own.

Firstly, the report implies that autopilot mode confusion played a role in the
late reaction of the pilots to the flight condition. They were expecting SRS
mode and got ALT* (for whatever reason) - they were expecting pitch protection
when there was none - they were waiting for something that wouldn't happen,
and they couldn't tell from the PFD.  Pete Mellor, in his article `CAD:
Computer Aided Disaster' and Robert Dorsett have noted that mode- or
control-law-confusion seems to have played a role in many of the A320
accidents as well.

Secondly, this airplane was loaded to within legal limits and was using thrust
appropriate to a go-around situation. There are US airports at which
commercial flights take place at which the missed-approach procedure requires
one to climb-and-maintain altitudes in the region of 2000ft. So, one might
consider the possibility that these three of the identified `causes' of the
accident were plausible, although maybe unusual, operating conditions.  The
airplane was pitched up by the copilot to 28 deg, in order (I would surmise)
to activate the automatic pitch protection mechanism, under conditions of
engine failure. Under these conditions, under autopilot control, the airplane
flew itself into an flight condition from which an experienced test pilot was
unable to recover in time. I wonder why more attention is not paid to this
feature of the accident?

The trim setting was singled out as a cause, but the report also says that the
accelerated rotation caused by this was controlled by the copilot, so I don't
see how it figures as a cause, unless it was seen as one-task-too-many.

For comparison and discussion in RISKS, I'd like to mention a possible point
of view different from that provided by Airbus [Ziegler interviews, Der
Speigel 15.8.94, RISKS-16.35, and Flight International, 17-23 Auf 1994, p4].
Namely: if the airplane had not crashed, seven more people would be alive -
but we also wouldn't have known that an A330 with full aft CofG is unable to
fly itself out of an engine-out-during-go-around situation if the
altitude-select on the AP is set at or near 2000ft and the pitch is slightly
above its 25deg limit of protection.

Is this computer-related?  I'm sure the A330 software will be changed.
If only because the Commission of Inquiry recommended it.

Peter Ladkin

ESORICS 94 Program

Yves Deswarte <>
Tue, 6 Sep 1994 14:17:01 +0100
Catherine Richards House, 16 Nelson Street, Southend-on-Sea, Essex, SS1 1EF.
Tel: (0702) 354020     Fax (0702) 354111

         (European Sympoisum on Research in Computer Security)

7TH - 9TH NOVEMBER, 1994

ESORICS-94 is organised by the IMA in co-operation with AFCET (creator),
BCS Computer Security Specialist Group, CERT-ONERA, AICA and GI

                         Provisional Program

Monday, 7th November, 1994

 9.15 -  9.30 a.m.      Introduction - Roger Needham and Gerard Eizenberg

 9.30 - 10.30 a.m.      Session 1 - Measures (Chair: Dieter Gollmann)

                        Valuation of Trust in Open Networks
                        T. Beth, M. Borcherding, B. Klein

                        Performance Requirements in Data Communication Systems
                        V. Zorkadis

11.00 - 12.30 p.m.      Session 2 - High Assurance Software
                        (Chair: John McLean)

                        Non-interference through Determinism
                        A.W. Roscoe, J.C.P. Woodcock, L. Wulf

                        Mechanical Proof of Security Properties
                        J.P. Banatre, C. Bryce, D. Le Metayer

                        Security through Types
                        C. O'Halloran, C.T. Sennett

 2.00 -  3.00 p.m.      Session 3 - Key Management I (Chair: Einar Snekkenes)

                        Designing Secure Key Exchange Protocols
                        C. Boyd

                        Robust and Secure Password and Key Change Method
                        R. Hauser, P. Jansson, R. Molva, G. Tsudik,
                        E. Van Herreweghen

 3.30 -  5.00 p.m.      Session 4 - Authentication (Chair: Emilio Montolivo)

                        Beacon Based Authentication
                        A. Jiwa, J. Seberry, Y.L. Zheng

                        Authentication via Multi-Service Tickets in the
                        Kuperee Server
                        T. Hardjono, J. Seberry

                        Oblivious Signatures
                        L. Chen


Tuesday, 8th November, 1994

 9.00 - 10.00 a.m.      Session 5 - Key Management II (Chair: Chris Mitchell)

                        A Model for Establishing Secure Channels in Open
                        U.M. Maurer, P.E. Schmid

                        On Strengthening Authentication Protocols to Foil
                        W. Mao, C. Boyd

10.00 - 10.30 a.m.      Session 6 - Invited Talk (presented by Chris Mitchell)

                        Security Research for the Financial Sector
                        H. Beker

11.00 - 12.30 p.m.      Session 7 - Digital Payment
                        (Chair: Jean-Jacques Quisquater)

                        Efficient Electronic Payment Systems Protecting Privacy
                        J.L. Camenisch, J.M. Piveteau, M.A. Stadler

                        The ESPRIT Project CAFE - High Security Digital
                        Payment Systems
                        J.P. Boly, A. Bosselaers, R. Cramer, R. Michelsen,
                        S. Mjolsnes, F. Muller, T. Pedersen, B. Pfitzmann,
                        P. de Rooj, B. Schoenmakers, M. Schunter, L. Vallee,
                        M. Waidner

                        Liability and Computer Security: Nine Principles
                        R.J. Anderson

 2.00 -  3.15 p.m.      Session 8 - Distributed Systems
                        (Chair: Peter Bottomley)

                        Implementing Secure Dependencies over a Network by
                        Designing a Distributed Secure SubSystem
                        B. d'Ausbourg

                        A Secure Medium Access Control Protocol:
                        Security vs Performances
                        P. Siron, B. d'Ausbourg

                        Distributed File Systems over a Multilevel Secure
                        Architecture, Problems and Solutions
                        C. Calas

 3.45 -  5.15 p.m.      Session 9 - Panel Session (Chair: Helmut Kurth)

                        Security Evaluation in Practice


Wednesday, 9th November, 1994

 9.00 - 10.30 a.m.      Session 10 - Access Controls
                        (Chair: Vijay Varadharajan)

                        On the Expressive Power of the Unary Transformation
                        R.S. Sandhu, S. Ganta

                        Privilege Graph: an Extension to the Typed Access
                        Matrix Model
                        M. Dacier, Y. Deswarte

                        A Consideration of the Modes of Operation for Secure
                        C. Robinson, S.R. Wiseman

11.00 - 12.30 p.m.      Session 11 - Database I (Chair: Catherine Meadows)

                        Mark-and-Sweep Garbage Collection in Multilevel Secure
                        Object-Oriented Database System
                        A. Ciampichetti, L. Mancini, E. Bertino

                        Decomposition of Multi-level Objects in an
                        Object-Oriented Database
                        N. Boulahia-Cuppens, F. Cuppens, A. Gabillon,
                        K. Yazdanian

                        Supporting Object-based High-assurance Write-up in
                        Multilevel Databases for Replicated Architecture
                        R. Thomas, R.S. Sandhu

 2.00 -  3.00 p.m.      Session 12 - Database II (Chair: Joachim Biskup)

                        Aggregation in Relational Databases:
                        Controlled Disclosure of Sensitive Information
                        A. Motro, D.G. Marks, S. Jajodia

                        Information Flow Controls vs Interference Controls:
                        An Integrated Approach
                        F. Cuppens, G. Trouessin

 3.00 -  3.15 p.m.      Conclusion - Roger Needham

GENERAL CHAIR: Roger Needham (University of Cambridge).

Miss Pamela Irving, The Conference Officer, The Institute of Mathematics
and its Applications, Catherine Richards House, 16 Nelson Street,
Southend-on-Sea, Essex, SS1 1EF.  Tel. (0702) 354020.  Fax. (0702) 354111.

::::: Yves Deswarte - LAAS-CNRS & INRIA - 31077 Toulouse (France) :::::
:::: - Tel:+33/61336288 - Fax:+33/61336411 ::::

Please report problems with the web pages to the maintainer