The Risks Digest

The RISKS Digest

Forum on Risks to the Public in Computers and Related Systems

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

Volume 23 Issue 31

Friday 9 April 2004

Contents

Chinooks again
Neil Youngman
Blackout computer failure analysis
Stephen Cohoon
Malware, auto-reply, and non-native languages
Drew Dean
Risks in Google's New "Gmail" Service
Lauren Weinstein
Risks in Network Solutions' domain information masking
Lauren Weinstein
Seeing the Light might just *not* show the right contamination
Bob Heuman
Re: Buffer overflows
Jon A. Solworth
Re: iAPX 432
Robert I. Eachus
Re: 4.6-million DSL subscribers' data leaked in Japan?
Curt Sampson
Re: News24's not-very-restrictive access restrictions
Curt Sampson
Yet another version of the Beagle social engineering
John Sawyer
REVIEW: "Cybersquatters Beware", Chantelle MacDonald Newhook
Rob Slade
Info on RISKS (comp.risks)

Chinooks again

<Neil <no.spam.for.n.youngman@ntlworld.com.die.spammers>>
Thu, 08 Apr 2004 06:42:36 +0100

The Royal Air Force bought 8 Chinook Helicopters for 259M pounds. The
helicopters were supposed to be in service 6 years ago, but problems with
radar systems, mean they can not fly in cloud. According to the BBC

  "The Chinooks were originally supposed to be in service in 1998 but radar
  systems and software developed under a separate contract would not fit in
  the cockpit, the report said."

http://news.bbc.co.uk/1/hi/uk/3606325.stm

  [The Chinooks are hiding in Chicrannies.  PGN]


Blackout computer failure analysis

<Stephen Cohoon <stephen.cohoon@voxpath.com>>
Fri, 9 Apr 2004 17:19:34 -0500

Kevin Poulsen of Security Focus reports on the analysis of the Great
Northeast Blackout of August 2003 in
 http://www.theregister.co.uk/2004/04/08/blackout_bug_report/
<http://www.theregister.co.uk/2004/04/08/blackout_bug_report/> .  Much is
made of the fact that the Energy Management System built by GE had a very
obscure, difficult to test race condition that caused it to fail during the
early stages of the blackout thus denying the operators at FirstEnergy much
of the information they needed to prevent or limit the blackout.

I would agree that this is a difficult bug to root out, but there are several
extremely important risks that need to be noted.

* The alarm system failed silently - the operators did not know that it had
  failed and was displaying stale data.

* The alarm queues backed up causing that host to fail - bad design that
  does not handle queue overflows.

* The backup alarm queue processor kicked in and quickly failed - bad design
  of an allegedly high-available system

These 3 points prove that the system suffered from much more than an obscure
race condition.  It had truly bad design problems.  Clearly, when GE created
an energy management system that processed alarms they would have benefited
from the experience of the telecom world where we have been building alarm
monitoring systems for years.  We made these mistakes years ago and they
should have known that.

The best part of the article:

  "But Peter Neumann, principal scientist at SRI International and moderator
  of the Risks Digest, says that the root problem is that makers of critical
  systems aren't availing themselves of a large body of academic research
  into how to make software bulletproof.

  ""We keep having these things happen again and again, and we're not
  learning from our mistakes," says Neumann. "There are many possible
  problems that can cause massive failures, but they require a certain
  discipline in the development of software, and in its operation and
  administration, that we don't seem to find. ... If you go way back to the
  AT&T collapse of 1990, that was a little software flaw that propagated
  across the AT&T network. If you go ten years before that you have the
  ARPAnet collapse.

  ""Whether it's a race condition, or a bug in a recovery process as in the
  AT&T case, there's this idea that you can build things that need to be
  totally robust without really thinking through the design and
  implementation and all of the things that might go wrong," Neumann says."


Malware, auto-reply, and non-native languages

<Drew Dean <ddean@csl.sri.com>>
Wed, 07 Apr 2004 11:51:54 -0700 (PDT)

I received the (excerpted, redacted) following piece of mail.

> From: ABCDEF HIJKLM <HIJKLM.ABCDEF@WXYZ.co.in>
> Subject: Out of Office AutoReply: good morning
>
> Hallo
> Your mails have not been received by me.
> Please send all your mail to HIJKLM.ABCDEF@NOPQRST.com.

My first thought was that this was a phishing expedition.  Interestingly
enough, all of the Received: headers in the mail seemed to be valid, and the
mail was routed through NOPQRST's mail servers.

Then it hit me: NOPQRST is a large, German multinational.  I'm not familiar
with WXYZ, although ABCDEF HIJLKM's name sounds Indian, so that's
reasonable.  It must have been the case that some piece of malware
(spreading its garbage) forged my name in the From: header.  "Hallo" is
typically German.  "Please send all your mail to" is fluent, close to
correct, but decidedly not idiomatic English -- easily explained if he's not
a native speaker.

So we have the confluence of malware, auto-reply agents, and non-native
speakers of human languages.

I decided not to forward all my spam to him. :-)

Meta-RISK: I'm not using string's of X's for redaction -- as usual, in an
attempt to get through spam filters.

Drew Dean, Computer Science Laboratory, SRI International


Risks in Google's New "Gmail" Service

<Lauren Weinstein <lauren@vortex.com>>
Fri, 02 Apr 2004 07:48:14 PST

Google (or ISPs) getting into the business of routinely scanning users'
e-mail for "interesting" keywords is of staggering import, even if the
reason is "merely" to insert ads (or spam control, for that matter, though
Google's plan to act as a massive long-term e-mail repository ups the risk
ante considerably over e-mail pass-through ISPs).

What would Google's legal responsibilities and actions be if they "stumbled"
across discussions of apparently illegal activity (everything from overdue
library books to adultery to murder...), or terrorism, or illicit
pornography?  Since they've apparently opened the surveillance box, it's
quite possible they'd be legally required to report everything that might
even potentially fall into questionable categories.

This of course would include all the false alarms that would be generated by
innocent messages that only looked suspicious but really weren't, not to
mention purposely faked messages spiked with likely nasty keywords to try
upset the system.  Even with the best of motives, do we really want Google
or ISPs becoming the commercial equivalent of Total Information Awareness?
We all want to prevent crime and terrorism, but is the creation of massive
surveillance machines in the guise of free e-mail services the proper way to
do so in our society?

And what of the proprietary information that will inevitably find its way
into Google's scannable e-mail treasure chest?  "Innocent" scanning could
reveal all sorts of goodies.  (I've thought in the past about all those new
product names and future trademarks that first drop into Google's logs when
initial searches are performed...)  Can we trust Google not to abuse this
potentially lucrative power?  For now the answer is probably yes, but market
forces make the future anything but certain.

Don't get me wrong.  I like Google -- a lot.  I think overall they've got a
good attitude, and a superb search engine (though the privacy implications
of their search logs have long been a matter of concern, as I noted).  But I
fear that they have not fully thought through the ramifications of their new
e-mail project, and how it can, even with the best of intentions, be rapidly
turned to the Dark Side.  That risk won't only result from Google's
decisions, but also from actions by government, lawyers, law enforcement,
courts, and even ISPs and Google's competitors.

E-mail is arguably the most sensitive form of Internet communications, and
deserves the highest possible levels of protection.  Mere trust or good
faith aren't enough.

In the classic (and highly recommended) satirical film "The President's
Analyst," the protagonists gradually come to the realization that every
phone call in the country is being tapped.  The 1967 film has been
prescient in numerous ways, and doesn't seem quite so funny anymore.

Centralized scanning of e-mail (even for ostensibly innocent commercial
purposes), the push for expanded surveillance of conventional and VoIP
telephone systems, and many other moves, together point towards a
future where all use of telecommunications is monitored through
close alliances of commercial enterprises and government, and
where encryption will be banned or tightly controlled.

Even if one assumes completely benign motives on the part of these
firms and governments today, what of the future?  Will the
incredibly powerful and pervasive monitoring infrastructures
now being woven always be in the hands of such trustworthy entities?

History suggests that we have a lot to worry about in these regards.


  [Incidentally, it is now being reported that Google's terms of service for
  Gmail (or whatever it ends up being called if there are really trademark
  problems) apparently say they can KEEP your e-mail even after you close
  your account!  That's what's being reported!]

Lauren Weinstein lauren@pfir.org  http://www.pfir.org/lauren +1 (818) 225-2800
Co-Founder, PFIR - People For Internet Responsibility - http://www.pfir.org


Risks in Network Solutions' domain information masking

<Lauren Weinstein <lauren@vortex.com>>
Mon, 29 Mar 2004 10:22:41 PST

Network Solutions (and other registrars in various ways) have begun offering
services to bypass ICANN's requirement for accurate domain holder
information being listed in WHOIS.

While the issues of WHOIS information accuracy and availability vs. privacy
are complex and controversial, NetSol's particular approach appears to
trigger a number of interesting legal questions.

For an extra charge, NetSol will mask the contact e-mail address with
an aliased address that changes at intervals, list their own po box
for the physical address, and list a phone number that leads to
a "no calls accepted" recording.

I do not see an obvious problem with their e-mail alias approach.  However,
their intercepting and opening of physical mail is a different matter, since
it makes it impossible for senders of certified mail to be assured that the
material ever reached the actual registrant, and of course privacy of such
mailings is lost.  If confidential legal materials were involved, the issues
could get very dicey.

Lack of an accurate phone number is of great concern.  In cases of network
disruptions (either intentional or not) often the only recourse to restore
normal services is to pick up the phone and call the person at the domain in
question.  Physical mail is too slow, and if systems are disrupted e-mail
often won't work.

One also must wonder if NetSol really wants to interject themselves into the
middle of legal and related communications involving spammers,
pornographers, and others with less than pristine motives for wanting to
hide their contact info -- the John Smith family who wants to protect their
home address is not the major issue.

Finally, what actions will ICANN take to enforce both the letter and spirit
of their rules in this regard, while these issues are being hashed out in
various policy forums?

Lauren Weinstein lauren@pfir.org  http://www.pfir.org/lauren +1 (818) 225-2800
Co-Founder PFIR <http://www.pfir.org>, Fact Squad <http://www.factsquad.org>


Seeing the Light might just *not* show the right contamination

<"R.S. (Bob) Heuman" <rsh@idirect.com>>
Tue, 06 Apr 2004 11:01:11 -0400

The real risks with the following "New Technology" are that it depends on
the light source remaining at the correct 'frequency' throughout its
lifetime AND somehow letting you know when it is NOT working correctly.
Until the bulb burns out you simply will not have a clue that it is not
working correctly should the light wave used shift frequency or the 'marker'
detection fail. There is no discussion of failure mode, unfortunately, so we
could all end up deep in what the product is intended to detect.

New technology could detect dirty hands, Associated Press, 5 Apr 2004
http://edition.cnn.com/2004/TECH/ptech/04/05/cleanhand.technology.ap/index.html

  With just a flicker of blue light, little Johnny's mother one day may know
  for sure whether her son washed his hands before dinner.  New light-scanning
  technology borrowed from the slaughterhouse promises to help hospital
  workers, restaurant employees -- one day, even kids -- make sure that hand
  washing zaps some germs that can carry deadly illnesses.  A device the
  size of an electric hand dryer detects fecal contamination and pinpoints
  on a digital display where on a person's hands more scrubbing is needed.


Re: Buffer overflows (Cowan, RISKS-23.30)

<"Jon A. Solworth" <solworth@cs.uic.edu>>
Tue, 06 Apr 2004 10:17:33 -0500

I'll take issue with Crispin Cowan's assertion that (RISKS-23.30):

  "Thus hardware semantics enforcement ended because the hardware people
  discovered that it was easier to do in software."

The issue is not whether there is hardware semantics enforcement.  All
modern architectures do have hardware enforcement via privilege modes, trap
instructions, and memory protection.  Rather the issue is of what form they
take, whether they provide only mechanism (and hence are policy neutral) and
what level of semantics they provide.

The problem of layering (whether to put the mechanism in the architecture,
operating system or application is exacerbated by two issues.  The first is
pushing to lower levels special purpose mechanisms.  The second is pushing
to higher levels issues we don't understand and making them more complex.
The latter is every bit as much of a problem as the former, and thereby
following Hoare's dictum:

 "I conclude that there are two ways to constructing software design: One
  way is to make it so simple that there obviously no deficiencies and the
  other way is to make it so complicated that there are no obvious
  deficiencies."

The higher levels of necessity depend on the lower levels for their
security.  So the question is how, not whether, the protection is provided
at the various levels.

Jon A. Solworth, Computer Science Dept., University of Illinois at Chicago
Chicago, IL 60607-7053  http://parsys.cs.uic.edu/~solworth 1-312-996-0955


Re: iAPX 432 (Cowan, RISKS-23.30)

<"Robert I. Eachus" <rieachus@comcast.net>>
Mon, 05 Apr 2004 19:48:51 -0400

>> Heck, software can always make up for hardware deficiencies, right?

> That's a perspective, but I disagree. Burroughs was not the only lab to
> experiment with language support on chip. Intel also tried this, with the
> i432. The result was that the RISC processors produced *crushing*
> performance wins over chips with complex semantics, due to the ability to
> heavily pipeline and thus ramp up the clock on simple instruction sets.

The death or rather non-success of the iAPX-432 had very little to do with
language support.  It had a lot more to do with the fact that customers were
not willing to pay for security--the iAPX-432 had very good hardware support
for security.

But the devastating decision was a decision not to fix a "bug" in the
initial version of the chip.  The net effect was to make every instruction
instruction owned by a user take two memory cycles longer to complete, and
every OS instruction took one more memory cycle when working with user data.
It is an instructive lesson in how software fixes for hardware problems may
be counterproductive.

The iAPX-432 was a capability based architecture.  All resources, in
particular memory were accessed through capability descriptors.  The chip
kept track of whether or not something was a capability, so they were not
easily forged.  However, in the final preproduction version of the chip,
there was a glitch.  As a result, the only way to expand the top-level
capability table was to do a hardware reset. (Ouch!) Of course, the
workaround that the developers used was to have a fixed set of top-level
descriptors, which pushed all user owned descriptors one level lower.

On a day when I was there for a job interview in the compiler department,
the decision was made to ship the current version of the chip, and fixed
this problem in a later mask release.  At lunch the discussion was not about
how bad this decision was, but where people were going to go when they left
Intel.  Needless to say, I didn't take the job.  (I probably should have --
the actual result of that decision by Intel was the founding of two
companies: Rational and Verdix.  Rational eventually bought Verdix, and was
then acquired by IBM.)

Why was the decision by Intel so bad?  Because it meant that almost anyone
designing products around the iAPX-432 made the same workaround decision.
Once you made that decision it didn't matter if 99.9% of the chips your
software ran on had the fix.  You had to design your software one way or the
other.  There was one company in England, High Integrity Products was the
name I think, who instead decided not to support the first batch of iAPX-432
chips.  They had quite a good business for years.  But that was it.  Well,
BiiN, a joint venture between Siemens and Intel was developing a trusted
operating system, and there was no reason for them to support the original
buggy chips.  But Intel discontinued the 432 and disbanded the company
before they shipped product.

It is interesting to speculate what would have happened if absent that
decision, the 432 became the future of computing.  Certainly most viruses
would never have existed.


Re: 4.6-million DSL subscribers' data leaked in Japan?

<Curt Sampson <cjs@cynic.net>>
Tue, 6 Apr 2004 14:16:03 +0900 (JST)

> Date: Sun, 21 Mar 2004 01:54:11 +0900
> From: Chiaki <ishikawa@yk.rim.or.jp>
> ...
> Really a good example of what one should NOT do in handling sensitive
> information and seeing it at an large ISP is rather disappointing and
> disgusting.

It may be disappointing and disgusting, but it's my opinion that this
is going to happen again and again. I have friends who work as computer
technicians in various companies here, and there are other companies
that keep similar personal information and take just as little care
with it. Not only have the managers been informed by tech folks time
and time again over the years that this risk exists, but even after
management has seen this whole Yahoo brouhaha they still have done
*nothing whatsoever* to make the information more secure. Honestly, not
one single step.

I myself in the past have been in a similar position and have tried to
convince managers that they should spend a little to reduce the level of
risk they were undertaking, and have always had little luck.

> This incidence is a really good example of what we should not do and
> many organizations seem to take this example seriously.
> In that sense, it is a good thing...

It would be interesting to find out how many organizations have taken
this example seriously, and how many have not. I can't think of any
way to do this, of course; everybody's going to say, in public, that
this can't happen to them. The only way to find out, if you're not part
of the organization itself, is to know someone in the right place in
the organization who's willing to tell you about it over a beer. And
needless to say, that kind of information, when handed to someone, is
probably not going further. I'm left with anecdotal evidence.

Possibly being able to buy insurance against these sorts of things
would help; insurance companies would presumably then be able to come
in and inspect for such situations, giving companies a direct financial
incentive to fix this stuff. And I would certainly be much more inclined
to use a company that had insurance that would pay me directly, say,
$250 should my personal information be compromised.

Curt Sampson  <cjs@cynic.net>   +81 90 7737 2974   http://www.NetBSD.org


Re: News24's not-very-restrictive access restrictions (RISKS-23.30)

<Curt Sampson <cjs@cynic.net>>
Tue, 6 Apr 2004 14:15:38 +0900 (JST)

> From: "Cody B." <cody@zone38.net>
> Thus, foreigners can read all of News24's articles in full, at absolutely
> no cost, *simply by turning Javascript off*.
> I don't even think I need to point out the many things that are utterly
> wrong with this.

It would help me if you would do so. I have often done things in a similar
style, after doing a risk and cost analysis. In those cases, the conclusion
that the programming team and the customer came to, after discussion, was
that a few users getting by the access restrictions was not a big problem,
and the cost of that was less than the cost of modifying the system so that
no users got by the access restrictions.

For a case like this, I can easily conceive of the following scenario:

  1. Modifying the back-end software to deal with this would be expensive
  because they would have to bring in an external developer to do the work
  and to retest. There's also additional risk in that should a bug be
  introduced, they have to call in the external developer again. (They use
  an external developer to do the development work on this part of the
  system because it doesn't change often enough to warrant hiring a
  full-time programmer to do the work internally.)

  2. Doing the javascript redirect is very cheap, because they already have
  a full-time HTML guy with the expertise to do this, and the system is
  designed to let him easily add such things to the requisite pages in the
  system.

  3. They currently have in place and use a system for analyzing their web
  logs that can tell them how many users are getting around the restriction.

None of the above points are particularly uncommon at many web-publishing
organizations, and I could certainly see this being a situation where you
might say, "the extra bandwidth charges for a few foreign users are not
going to be substantial and there seems to be no other risk of us suffering
from something we're not already suffering from now; let's do it the cheap
way, and only re-implement it the more expensive way when we see that more
than a certain percentage of our foreign users are getting around the
restrictions."

I didn't see any indication in your message that you have sufficient
information to do the analysis above and have done this analysis, and found
that this javascript course was not the best one to take.

The RISK here? Perhaps it's assuming you have all the information you
need to do an analysis of whether something is a smart or stupid move.

Curt Sampson  <cjs@cynic.net>   +81 90 7737 2974   http://www.NetBSD.org


Yet another version of the Beagle social engineering

<=?iso-8859-1?q?John=20Sawyer?= <jpgsawyer@btopenworld.com>>
Tue, 6 Apr 2004 08:52:10 +0100 (BST)

Following up from my recent report on 16 Mar in RISKS-23.27 of social
engineering to get the Beagle virus onto people's systems, here is yet
another attempt.

>Dear user of Btopenworld.com gateway e-mail  server,

> Some of our clients complained about the spam (negative e-mail content)
> outgoing from your e-mail account.  Probably, you have been infected by a
> proxy-relay trojan server. In order to keep your computer safe, follow the
> instructions.
>
> Further details  can  be  obtained  from attached file.
>
> Have a good  day,
>    The Btopenworld.com team
>
> http://www.btopenworld.com

This time they interestingly attempt to scare people into opening the
attachment by accusing them of already having had their computer compromised
and being responsible for spam. An old con trick but not one I have seen in
this context before. Another interesting feature is that the message doesn't
seem to have any gross grammatical errors this time something that seemed to
be standard in these things before.

John Sawyer, University of Portsmouth


REVIEW: "Cybersquatters Beware", Chantelle MacDonald Newhook

<Rob Slade <rslade@sprint.ca>>
Wed, 7 Apr 2004 08:19:07 -0800

BKCYBSQT.RVW   20031118

"Cybersquatters Beware", Chantelle MacDonald Newhook, 2002,
0-07-090579-7, U$19.95/C$24.99
%A   Chantelle MacDonald Newhook chantelle@disputewinners.com
%C   300 Water Street, Whitby, Ontario   L1N 9B6
%D   2002
%G   0-07-090579-7
%I   McGraw-Hill Ryerson/Osborne
%O   U$19.95/C$24.99 905-430-5000 800-565-5758 fax: 905-430-5020
%O  http://www.amazon.com/exec/obidos/ASIN/0070905797/robsladesinterne
  http://www.amazon.co.uk/exec/obidos/ASIN/0070905797/robsladesinte-21
%O   http://www.amazon.ca/exec/obidos/ASIN/0070905797/robsladesin03-20
%P   290 p.
%T   "Cybersquatters Beware"

The introduction talks about branding, and the fact that companies can
do it very inexpensively on the Internet.  Chapter one explains
trademarks, as well as the domain name system.  The Uniform Domain
name dispute Resolution Process (UDRP) is outlined in chapter two
(along with other similar mechanisms), and key elements necessary to
winning a dispute are noted.

Successive chapters present a number of cases involving different
types of principals and principles: companies (in three), institutions
and individuals (in four), celebrities (five), sex (six), complaints
and comments (in seven), generic names (eight), and an amalgam, in
chapter nine, of airlines, banks, wineries, and other companies that
have not prepared for the disputes.  Chapter ten deals with the
process of going to court with domain name disputes.  Trends and
indications in decisions are reviewed in chapter eleven.

The book does provide a good compilation of advice on a complex and
poorly understood topic.  There is one proviso: the text frequently
makes the point that the race is not always to the justified, nor the
legal battle to the prepared.  However, as current wisdom has it, the
prepared side is the one to bet on.

copyright Robert M. Slade, 2003   BKCYBSQT.RVW   20031118
rslade@vcn.bc.ca      slade@victoria.tc.ca      rslade@sun.soci.niu.edu
http://victoria.tc.ca/techrev    or    http://sun.soci.niu.edu/~rslade

Please report problems with the web pages to the maintainer

Top