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 55

Thurs 30 September 2004


Federal Judge Strikes Down Part of PATRIOT Act
CDT via Monty Solomon
Nationwide Radio Shack outage
George Coulouris
Georgia's computer systems down for 16 hours
Bob Harbort
Voter-verified paper trails vs. Internet voting
Lauren Weinstein
Swiss tout success of Internet-voting test
Dutch Internet elections
Erling Kristiansen
Gov. Schwarzenegger signs CA paper trail bill into law
Kim Alexander
JPEG/GDIplus vulnerability
Rob Slade
Realtime keyword voice recognition... not just for the NSA anymore
Danny Burstein
Software that knows your every move
Burt Helm via Monty Solomon
The risks of zero feedback
Ian Chard
Free ISPs safe?
Dan Jacobson
Fraud e-mail detector risks
Danny Lawrence
Re: Java programs at risk from decompilers
Steve VanDevender
Info on RISKS (comp.risks)

Federal Judge Strikes Down Part of PATRIOT Act

<Monty Solomon <>>
Thu, 30 Sep 2004 00:36:06 -0400

Date: Wed, 29 Sep 2004 14:42:47 -0400 (EDT)
Subject: CDT Headline: Federal Judge Strikes Down Part of PATRIOT Act
List-Archive: <>

A federal judge today found unconstitutional a part of the USA PATRIOT Act
that allows federal law enforcement officials to obtain confidential
financial records without a court order or other safeguards.  The lawsuit,
brought by the ACLU, challenged the use of so-called "National Security
Letters," a type of administrative subpoena power that was expanded by the

For more on the USA PATRIOT Act:

For more information on the ACLU lawsuit [offsite]:

Nationwide Radio Shack outage

<George Coulouris <>>
Mon, 27 Sep 2004 08:56:26 -0400

On Sunday, September 26, I attempted to exchange a $21.99 cable for a $26.99
cable at my local Radio Shack. The manager informed me that all 7000 Radio
Shack stores nationwide were unable to process transactions due to a
computer outage. I asked if I could simply leave the $5 difference in cash
and let the store sort things out when they were back online, but I was told
that this was not possible and to please come back in a couple hours.

Georgia's computer systems down for 16 hours (Re: RISKS-23.52)

Thu, 30 Sep 2004 12:30:45 -0400 (EDT)

A note in RISKS-23.52 about the state computers in Illinois being offline
for an hour prompts me to send in the following. Georgia may be last in
piddlin' stuff like SAT scores and teacher salaries, but we know how to have
*major* downtime!

  "A 16-hour shutdown of the state government's computer networks Tuesday
  could have been prevented if officials had heeded an earlier warning that
  the computers needed replacement batteries.  The state's computers froze
  for the entire business day Tuesday, shuttering tag offices and delaying
  court-ordered child support payments to 516,000 Georgia children, among
  other problems.  The shutdown was caused by a power outage at Georgia
  Power Co. as the remnants of Hurricane Frances swept through the state.
  But the computers never should have been running on electricity. An April
  report by a state consultant noted that batteries that normally power the
  computers had gone bad and two backup generators had failed.

  Officials with the Georgia Technology Authority told The Atlanta
  Journal-Constitution that they learned of that consultant's report just
  last month, and didn't have time to implement its recommendations before
  this week's bad weather."

I particularly like the part about "the computers never should have been
running on electricity." White lightnin', maybe?

Bob Harbort, Southern Polytechnic State U., CS/Software Engineering
1100 S. Marietta Pkwy. Marietta, GA 30060-2896 1.678.915.7405

Voter-verified paper trails vs. Internet voting

<Lauren Weinstein <>>
Wed, 29 Sep 2004 11:43:16 -0700

With California just wisely enacting a secure, voter-verified paper trail
law relating to electronic voting machines, many other states following a
similar path, and related federal legislation also being considered, it
appears that this methodology is likely to become a fixture in many, most,
or even all U.S. jurisdictions where touch-screen or other e-voting machines
are in use.

However, I haven't seen discussion of a likely side-effect of this welcome
trend -- its impact on proponents of voting over the Internet.  If a secure,
voter-verified paper trail is being required for e-voting machines operating
in the theoretically "secure" environment of polling places, it's quite
possible that we're finally putting a stake in the heart of the extremely
ill-advised and highly risky idea of Internet-based voting, even if the laws
regarding paper trails may not address this issue specifically.

Keep in mind that typical voter-verified paper trail systems permit the
voter to inspect the receipt but not to handle or remove it -- the receipt
must stay under the control of the voting authorities at all times.  These
twin requirements cannot be simultaneously met in remote Internet voting
situations (e.g., people voting from their homes or offices, which are the
big "selling points" for those persons pushing Internet voting).

No doubt the proponents of Internet voting will suggest all manner of
bizarre schemes involving encoded receipts that could be physically mailed
back to voting authorities or other similar completely impractical ideas.

But the bottom line is that if a secure, voter-verified paper trail is
needed for e-voting machines -- and it is -- then Internet voting should be
considered to be dead on arrival for the foreseeable future at least.

Lauren Weinstein
+1 (818) 225-2800

    [Incidentally, the October issue of the *Communications of the ACM*
    has a special section with 8 papers devoted to voting systems.  PGN]

Swiss tout success of Internet-voting test

<"NewsScan" <>>
Tue, 28 Sep 2004 09:12:02 -0700

Switzerland has declared Internet voting a success after being used without
problems in a national referendum in which 2,723 people in four Geneva
suburbs visited a special Web site to cast their votes on various
issues. Swiss officials says the test has proven their systems are robust
and secure, but e-voting critic Avi Rubin of Johns Hopkins University is
unconvinced: "Just because nobody attacked a referendum that involved 2,723
people does not mean that it was secure. When these trials are viewed as
successful and justify more in-depth electronic voting, eventually there
will come a point where it will be worth someone's while to attack the
election." Geneva's e-voting system uses software developed by local
election authorities with the Swiss office of Hewlett-Packard Co. and the
Geneva-based online security firm Wisekey.  [AP/*USA Today*, 27 Sep 2004;
NewsScan Daily, 28 Sep 2004]

Dutch Internet elections

<Erling Kristiansen <>>
Sun, 26 Sep 2004 21:41:53 +0200

The Netherlands is currently holding the election for the Regional Water
Management Boards (my translation of "Waterschappen"). One can vote by mail
or by Internet.  The latter attracted my curiosity, and I poked around the
'net a bit to see what people thought about the idea.

It appears that a test election was held in order to test the procedure and
get some feedback from test voters. An often quoted feedback was that "Only
26% of the test voters expressed concern about the possibility of
fraud". ONLY 26%?? This response seems to be interpreted as a vote of
confidence for the system.

Another nugget: "The secrecy of the vote is guaranteed. The relationship
between the voter identity and his login code is removed from the file
before the votes are counted".

The FAQ also has an interesting statement. An independent body (TNO) has
investigated the security of the voting method. They concluded that "Voting
by Internet is not less safe than voting by mail or phone". This formulation
implies that the procedure is actually not very safe, and they know it.

I cast my vote today. And my wife's!  With her permission, but I could
equally well have voted for her without her knowing about it. All you need
in order to vote is a couple of codes contained in a letter delivered by
post. Since I happened to be the one emptying the mailbox yesterday, I
retrieved my own ballot, my wife's and my daughter's.  And what about mail
delivered to the wrong address, that happens quite regularly. But not today,
so, sadly, I was unable to vote on behalf of any of my neighbours.

Slightly to my surprise, the voting web site worked well in Netscape, except
that Netscape aborted just after I had completed the second vote.

At the end of the voting process, you get a code, consisting of a total of
40 hex digits in 3 sub-fields. Allegedly, this code can be used, after the
election has closed, to "check whether the vote was counted".  If you do not
take note of these codes here and now, they are lost for good, there seems
to be no way to retrieve them later.  The page displaying the codes posts
the warning "Keep these codes secret, your vote can be derived from them".

It seems to me that there is no way to guarantee that "the vote was
counted", only that it was registered somewhere. You have to take their word
that this is actually the file that is counted, that nobody messed with it,
and that counting is done correctly and honestly.

The Water Management Board election is not a very high-profile election,
many show little interest in what these boards are doing. You vote for
individuals, not for political parties, and the campaign preceding the
election is not very intense, possibly because no party politics is
involved. So it may not be very fraud-sensitive. But I am afraid that if
this type of elections is declared a success, the technology may show up in
more important elections.

Gov. Schwarzenegger signs CA paper trail bill into law

Tue, 28 Sep 2004 10:47:00 -0700

Great news to share -- California Governor Arnold Schwarzenegger signed SB
1438 into law.  This bill, co-authored by Senators Ross Johnson (R-Orange)
and Don Perata (D-Alameda) requires there to be a voter verified paper
record to back up every electronic ballot cast in California by 2006 Primary

California is the first state in the nation where paperless, electronic
voting systems have been widely deployed that is requiring by law that the
machines be retrofitted or replaced.  With the enactment of SB 1438,
California continues to lead the nation on electronic voting reform.

Two other states -- New Hampshire and Oregon -- have laws that mandate the
use of voting systems that allow for manual recounts.  Illinois passed a law
that requires a voter verified paper trail for e-voting machines once that
state begins purchasing them.  Five Secretaries of State have elected to
implement the voter verified paper trail.  The first to do so is Dean Heller
of Nevada, who is implementing the paper trail this election season.  In
addition, the Secretaries of State in Washington, Missouri, California, and
Ohio will require the paper trail by 2006.

Although paper trail legislation was introduced in as many as 20 states this
year, it appears that California is the only state so far to enact a paper
trail law.  SB 1438 essentially codifies California Secretary of State Kevin
Shelley's November 2003 and April 2004 security directives, and advances the
deadline for implementing the paper trail by one election.  Under the
Secretary of State's orders, California would have the paper trail by the
November 2006 election.  SB 1438 ensures the paper trail will be in place
for the 2006 Primary.

The new law also prohibits the Secretary of State from certifying any new,
paperless electronic voting systems after January 1, 2005, and prohibits
counties from purchasing such systems after January 1, 2006.

Proclaimed "dead" just last month, SB 1438 was brought back to life by its
authors in the 11th hour of the legislative session, and sailed out of the
Legislature on unanimous votes of both houses.

For more information about SB 1438, visit
  bill_number=sb_1438&sess=CUR&house=B&author=johnson  [SPLIT URL]

For past CVF-NEWS updates on SB 1438, visit .

-- Kim Alexander, President, California Voter Foundation,, 916-441-2494

Contact the California Voter Foundation: 530-750-7650   U.S. Mail -  503 4th Street, Suite 6, Davis, CA 95616

JPEG/GDIplus vulnerability

<Rob Slade <>>
Sun, 26 Sep 2004 15:03:55 -0800

If you have not been living under a rock (in security terms), you will
likely have heard something about the GDI+ vulnerability in the past few
days.  JPEGs and other files that may be handled in the same way are now
potentially "dangerous" data files.

In 1994 a graphics file was spread via Usenet that contained oddities in the
header, and at about the same time a virus warning hoax was created that
warned of a viral JPEG file.  Neither of these was, in fact, related to
actual malicious software, but I did some study on the subject and found
header structures in both formats that could, potentially, have been used as
malware vectors, under certain conditions.

The specifics of the current JPEG/GDI+ vulnerability are very difficult to
obtain, even when you have copies of the various "exploits" that have been
released.  However, it does seem to be simply your common or garden buffer
overflow.  As I write I am not aware of any specific exploits that have been
released with the intent to use them maliciously.  However, given the number
of "exploit" samples that have been released I dare say that it will not be
long before we see the real ones come out.  It is unlikely that viruses will
be created using this vulnerability, but it is quite probable that viruses
will be created that carry graphics files (likely pornographic) that will
use the vulnerability to open links to malware on Web sites, or simply open
backdoors on machines for exploitation and amalgamation into botnets of
various types.

Microsoft security bulletin MS04-028
has some links that, if you manage to follow them all the way through, will
lead you to a patch.  Affected systems use certain versions of the
gdiplus.dll file.  The most widespread of the affected versions of the file
come with Microsoft Windows and Office, 2003 and XP versions.  Other
Microsoft, and other, products also have vulnerable versions of the file.

The file is fairly ubiquitous.  I've got eleven copies (and two compressed
copies) of five different versions of gdiplus.dll on my machine.  The
Microsoft site does provide details of which version numbers are vulnerable
or not--but no information about file sizes or dates that might allow you to
determine which versions are which.  If you follow links through from that
page there is also a "detection" tool--but it only tells you that you *are*
vulnerable, rather than identifying specific instances.

SANS also has provided a scanning tool, at
(Actually two, a GUI version and a command line version.  The GUI version,
as provided, seems to want a disk in drive F:, but if you tell it to
continue seems to function.)  This tool identifies which versions are
vulnerable and which are not, and also scans other filenames which are, in
fact, renamed copies of the gdiplus.dll file, such as:

   Version: 5.1.3097.0 <-- Vulnerable version
C:\Program Files\ArcSoft\Software Suite\PhotoImpression
   Version: 5.1.3097.0 <-- Vulnerable version
C:\Program Files\Common Files\Microsoft
   Version: 11.0.6360.0
C:\Program Files\Common Files\Microsoft Shared\VGX\vgx.dll
   Version: 6.0.2800.1106 <-- Possibly vulnerable (Win2K SP2 and
   SP3 w/IE6 SP1 only)
C:\Program Files\Microsoft Office\OFFICE11\GDIPLUS.DLL
   Version: 6.0.3264.0

Banning JPEGs is unlikely to be effective as a security measure.  Untrained
users will probably not know how to turn off the relevant functions, or be
willing to so "cripple" their Web browsing.  In any case, graphics files of
various types can be renamed, and Windows will still identify them from
internal structures, and run them through GDI+.

Microsoft has provided some new patches (patches for Office and Windows
apparently have to be installed separately), and others will possibly do so
as well.  It may be difficult to find the appropriate patches for all
applications.  One would assume that all versions of gdiplus.dll could
simply be replaced by the latest (safe) version, but, knowing the industry,
one would probably be wrong.    or

[Later update, Tue, 28 Sep 2004:]

a) I've seen at least one actual use of the exploit in what appears to be a
   malicious situation.

b) There are serious questions about whether the Microsoft updates (for both
   Windows and Office) actually work.

Realtime keyword voice recognition... not just for the NSA anymore

<danny burstein <>>
Mon, 27 Sep 2004 13:04:30 -0400

"Phone a call centre and you are likely to spend ages on hold listening to
canned music - and then find the operator cannot find the information you
need.  But an artificial intelligence system that hunts down the required
information is aiming to slash the time people waste this way.

"Using a mixture of speech recognition and search engine technology, the
system, being developed by IBM, will trawl a call centre's databanks for the
information a customer wants and present it to the operator before the
caller has finished explaining what they want. By giving operators rapid
access to the right information, calls will be dealt with faster.

"The system works by listening in to the conversation and identifying
keywords spoken by the customer....

Risks: aside from the commercial aspects, how long before these programs
start looking for any "disturbing" phrases in more and more communications,
and then immediately redflagging the call and forwarding the info to Law
Enforcement Agency of Your Choice?

Software that knows your every move

<Monty Solomon <>>
Mon, 27 Sep 2004 02:55:33 -0400

Source: Burt Helm, *Business Week*, 23 Sep 2004

It's called Worklenz, and it can be a powerful management tool for tracking
projects and people -- or a scary Big Brother

Look busy -- Worklenz is watching. Designed by privately held
information-technology company Métier in Washington, D.C., Worklenz is
software designed to help companies manage large projects and maximize
efficiency. But unlike an enterprise resource program, which tracks a
company's inventory, invoices, and assets, Worklenz tracks workers -- what
they do, when they do it, and how long it takes.

And it's spreading fast. Métier says it has been profitable for just over
two years and has won contracts with Lockheed Martin ( LMT ), BMW, Northrop
Grumman ( NOC ), and the U.S. Agriculture Dept. In the next week,
BusinessWeek Online has learned, Métier will announce a contract with the
FBI to manage all of the bureau's IT-related projects.

In its essence, Worklenz uses an extreme form of micromanagement to help a
company make broad decisions. The program can sync with each employee's
Microsoft Outlook e-mail account, Microsoft Project scheduling software, and
his or her PeopleSoft timesheet, to let a boss see everyone's schedules,
what tasks they're working on, and how soon each employee will complete his
or her work.  ...

The risks of zero feedback

<Ian Chard <>>
26 Sep 2004 17:08:03 +0100

Background: a Spanish bank issues its customers with a passbook instead ofan
ATM card on certain types of account (like UK building societies used to
do).  Its ATMs have a wide slot so that the passbook, open at a certain
page, can be inserted.  The book is about the size of a passport and
contains a history of transactions on the account.

From a hotel window overlooking such an ATM, I recently watched a woman
trying to use the service for the first time.  She opened the book and
inserted it correctly, and the machine pulled it completely inside.
However, the screen stayed in what pinball machine designers would call
"attract mode": it continued idly to display rolling adverts as if nothing
had happened.  Thirty seconds passed before a bank employee noticed her
waving through the window and beckoned her inside.  A few seconds later, the
machine spat out the book saying it couldn't be read.  It sat hanging out of
the slot for a full minute before the woman came out and retrieved it.

The risk is obvious, and I've also noticed in the last year or so that newer
ATMs in the UK also exhibit this behaviour, leaving the customer wondering
whether it even noticed that their card has been inserted.  It's so easily
mitigated ("Please Wait") that I find it hard to understand how these things
ever passed their testing phase.

Free ISPs safe?

<Dan Jacobson <>>
Sun, 26 Sep 2004 07:30:15 +0800

Here in Taiwan there are tons of free ISPs. One just dials up with the
username and passwd from their ads. But how long before one of them starts
snooping the data they pass for fun and profit?

Fraud e-mail detector risks

<Danny Lawrence <>>
Wed, 29 Sep 2004 16:37:10 -0400

Citibank has an e-mail account to send fraud/Phishing messages to,
presumably so that their security people can track down the perpetrators.  I
received an obvious (both to me and to my Spam catching software -- it was
in my "probable spam" file) phishing message so I dutifully forwarded it to
their address:

Later on I get a delivery failure: Error transferring to; SMTP Protocol Returned a Permanent Error 554
5.7.1 Virus present: Phish-BankFraud.eml

I suppose it is a good thing that Citibank has something that detects
fraudulent e-mails, but wouldn't it be better if they didn't use it on
messages coming into their fraud account?

Re: Java programs at risk from decompilers (O'Marcaigh, RISKS-23.54)

<Steve VanDevender <>>
26 Sep 2004 01:43:17 -0700

The risk here seems not to be the possibility that compiled programs can be
decompiled (I've heard of decompilers for various languages and system
architectures), but that there are people who wrongly think that a compiler
somehow conceals the algorithms used in a program.  Since a compiler's job
is to translate a source specification of an algorithm into a
machine-executable form, the compiler's job is to exactly preserve the
programmer's "work and intellectual property" in that sense.  The
machine-executable form might be less human-readable, but it is still
susceptible to analysis and reverse-engineering.  Compiled machine code is
often easy to decompile, even by hand, because compilers tend to generate
code with more conventional structure than human-written machine code.

> This new version would be an exact copy of the original program, but
> with a malicious payload.

How can such a program be "an exact copy" while having also been modified
"with a malicious payload"?  If it's been modified, this modification will
be easily detectable by comparison with the original.

At one time, "binary patching" (direct modification of machine code) was
commonly used for fixing bugs in code which could not be conveniently
recompiled or reassembled.

Apparently the now-ubiquitous use of compiled high-level languages is
causing people to forget what compilers actually do, and ignorance of the
details of low-level machine code leads to the erroneous idea that a
compiled form of a program is somehow "protected" from modification or
exposure of its inner workings.

  [RISKS received a similar note from Russ Perry Jr., and another reminder
  of the futility of security by obscurity from Dave Minter.  Scott Nicol
  cited two Java decompilers
  and observed that .Net code can also be decompiled.  He also added
    "Think how much easier Y2K fixes would have been if 1960's Cobol
    programs compiled into a form that was decompilable."

  See a somewhat related article that has just appeared:
    Huaiqing Wang and Shuozhong Wang,
    Cyber Warfare: Steganography vs Steganalysis,
    *Communications of the ACM*, 47, 10, October 2004, 76--82.   PGN]

Please report problems with the web pages to the maintainer