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 25 Issue 88

Saturday 26 December 2009

Contents

Insurgents Hack U.S. Drones
PGN
Another user interface fatal accident in Afghanistan
Mark Thorson
Security in the Ether: Cloud Computing? Or "Swamp" Computing?
Lauren Weinstein
HP's facial-recognition can't recognize black people's faces
Randall Webmail
Alert: Twitter apparently hacked
Lauren Weinstein
Silent Hybrid Nearly Causes Carbon Monoxide Poisoning
Bob Gezelter
UAL: Another risk of weather for computer based systems
Jared Gottlieb
When the human model doesn't match the system model
Sean W. Smith
Disconnects between the Real World and Cyberspace
Bob Gezelter
Obscure GPS problems not just in remote areas
Jeremy Epstein
On the Road with a GPS System
Gene Wirchenko
GPS ads for captive bus riders
jidanni
Cruise control failed to disengage
Steve Cody
Re: LED Traffic Lights are efficient but cannot melt away snow
John Johnson
Info on RISKS (comp.risks)

Insurgents Hack U.S. Drones

"Peter G. Neumann" <neumann@csl.sri.com>
Thu, 17 Dec 2009 16:28:57 PST

Militants in Iraq have used $25.95 off-the-shelf software to intercept live
video feeds from U.S. Predator drones, potentially providing them with
essence, the Predator video link to the ground station is unencrypted.
Insurgents are sniffing the link and reconstructing the video.  The
U.S. government has known about this flaw for a decade, but ignored it,
offering the usual litany of problems: encrypting the link would delay the
program, cost more, and complicate operations.  ``But the Pentagon assumed
local adversaries wouldn't know how to exploit it, the official said.''  The
stolen video feeds also indicate that U.S. adversaries continue to find
simple ways of counteracting sophisticated American military technologies.
[Source: Siobhan Gorman, Yochi J. Dreazen and August Cole, $26 Software Is
Used to Breach Key Weapons in Iraq; Iranian Backing Suspected, Wall Street
Journal, 17 Dec 2009, front page; PGN-ed] The myth of security by obscurity
strikes again.  http://ebird.osd.mil/ebfiles/e20091217723006.html

I've seen arguments that key management would be too difficult to manage,
although some reports say that efforts are now underway to encrypt.  Other
arguments are that uncleared soldiers need access to the video streams, but
that seems to confuse "encryption" and "classification".  Don't forget that
the former does not require the information to be classified for
confidentiality or that it is not important for integrity!  Also, the new
Reaper drones cost over $10 million each and have the same vulnerability.

Jeremy Epstein noted that *WiReD* reports that it's not just drones, but
also regular combat aircraft - although it's harder to intercept the
(unencrypted) signal from the regular planes.  The talk I've heard today
that "it's not such a big deal" seems odd - if you were an insurgent, having
a few minutes warning that there's a drone coming your direction would be
useful data.
http://www.wired.com/dangerroom/2009/12/not-just-drones-militants-can-snoop-on-most-us-warplanes/

Also, see Bruce Schneier's essay on the topic:
http://www.wired.com/politics/security/commentary/securitymatters/2009/12/securitymatters_1223


Another user interface fatal accident in Afghanistan

Mark Thorson <eee@sonic.net>
Tue, 15 Dec 2009 20:22:41 -0800

The circumstances of this incident seem very similar to the incident a few
years ago in which changing batteries cleared the offset between the
observer's GPS position and the target position on his Garmin.
http://www.dailymail.co.uk/news/article-1236087

When the target position is cleared, it would probably be better to
initialize it 100 meters to the east or something, rather than right on top
of the observer position.


Security in the Ether: Cloud Computing? Or "Swamp" Computing?

Lauren Weinstein <lauren@vortex.com>
Thu, 24 Dec 2009 11:04:16 -0800

Security in the Ether: Cloud Computing? Or "Swamp" Computing?
  [From NNSquad, Network Neutrality Squad, http://www.nnsquad.org]

An important article worth reading:

http://bit.ly/4uYabf  (MIT Technology Review)

My personal "thumbnail" view on this is that:

a) Cloud Computing" holds enormous promise.

b) Most of the key security and other operational issues associated with
   cloud computing are solvable, including aspects of pervasive encryption
   that would protect cloud computing clients from potential snooping by
   theoretically postulated unscrupulous cloud service providers.

c) The financial and intellectual resources (including basic policy
   analysis) required to understand and solve these problems on an *a
   priori* basis, rather than on an "after there's a mess" reactive basis,
   are in general insufficiently emphasized and deployed.

d) Given (c), not all of the current rush to cloud computing on today's
   widely available platforms can necessarily be justified as wise,
   particularly where sensitive and/or privacy-critical data is involved.

Or in other words, Cloud Computing can be a Really Good Thing if done
right, but let's not get the cart in front of the horse.


HP's facial-recognition can't recognize black people's faces

Randall Webmail <rvh40@insightbb.com>
December 21, 2009 1:29:12 PM EST

  [From Dave Farber's IP]
This is awkward. It appears that HP's new webcams, which have facial-
tracking software, can't recognize black faces, as evidenced in the above
video. HP has responded:

> We are working with our partners to learn more. The technology we use is
> built on standard algorithms that measure the difference in intensity of
> contrast between the eyes and the upper cheek and nose. We believe that
> the camera might have difficulty "seeing" contrast in conditions where
> there is insufficient foreground lighting.

> HP Face-Tracking Webcams Don't Recognize Black People - Hp - Gizmodo
> (21 December 2009)
http://gizmodo.com/5431190/hp-face+tracking-webcams-dont-recognize-black-people
http://snipurl.com/tsfli

Archives: https://www.listbox.com/member/archive/247/=now


Alert: Twitter apparently hacked

Lauren Weinstein <lauren@vortex.com>
Thu, 17 Dec 2009 22:38:41 -0800

Twitter has apparently been hacked.  Invalid security certificate for wrong
site on the Twitter https: home page, Iranian Cyber Army hack page on main
twitter.com.  This looks much like an SQL injection exploit I've dealt with
in the past, but I don't know for sure at this point if the actual Twitter
infrastructure has been hacked or if this is a DNS hack.

Presumably this won't last long, but more info when available.

Date: Fri, 18 Dec 2009 09:47:59 -0800

Twitter has not officially released details on last night's hacking-related
outage of their Web site, other than to state that it was (as many of us
suspected) a DNS-related attack.

There are some other details floating around unofficially.  Twitter's DNS
services are provided by Dyn Inc.'s Dynect Platform.  Dyn is insisting that
their systems were not compromised and that nobody accessed Twitter's DNS
data without appropriate (login) credentials.

This suggests (but again, this is *not* confirmed) that Twitter's account on
Dyn was somehow itself compromised, possibly through "social engineering" or
other techniques that resulted in the attackers gaining login access to the
Twitter account on Dynect, allowing them to change the associated DNS data.
(From Dyn's standpoint, this could still be considered to be "appropriate
login credentials.")

It goes without saying that the "Iranian Cyber Army" hack page is almost
certainly a fraud, and there are no indications that Iran actually had
anything to do with this attack (breathless statements blaming Iran being
made by some media points notwithstanding).  By the way, I've seen this
exact page resulting from various bot-based, non-DNS attacks in the past.

Presumably more "official" statements about what transpired will be
forthcoming at some point, after the finger-pointing slows down a bit.

Of course this once again demonstrates the fragility of DNS, but that's
hardly a headline news revelation at this stage of the game.

Date: Fri, 18 Dec 2009 12:14:27 -0800

Now confirming [ Ref: http://www.nnsquad.org/archives/nnsquad/msg02460.html ]
that the Twitter DNS diversion last night was the result of someone using
Twitter's own login credentials to change DNS data at Dyn's site,
according to Dyn's CTO:

http://bit.ly/80Ve4Y  (Wired)

So as suspected, this was not a "sophisticated" attack (e.g., DNS cache
poisoning) but rather a conventional login attack.

It is interesting to consider that apparently a single username/password
pair was able to take Twitter's entire Web site effectively offline
globally.

At the very least one would hope that more advanced account control
mechanisms (e.g., certificate-based access authentication) would be employed
with critical accounts for organizations at this level.


Silent Hybrid Nearly Causes Carbon Monoxide Poisoning

Bob Gezelter <gezelter@rlgsc.com>
Sat, 26 Dec 2009 08:07:53 -0500

The Decemeber 27, 2009 Sunday New York Times Magazine (p. 8) contains a
Letter to the Editor from Liz Cantarine entitled "Artificial Car Noises".

Ms. Cantarine describes how she accidentally left her hybrid automobile
running in her garage. Apparently, when returning from a shopping trip, she
became preoccupied with the packages in the trunk, and forgot to turn off
the car. The car continued running, filing the garage space with fumes.

It is an interesting problem. Synthetic noise generators are being required
due to a hazard to visually impaired pedestrians, but this is the first
report I have seen describing a danger to owners and their families from
silent cars.


UAL: Another risk of weather for computer based systems

jared gottlieb <jared@netspace.net.au>
Sun, 20 Dec 2009 15:51:55 -0700

From the United Airlines website 20 December 2009

Weather-related delays and cancellations resulted in a significantly
higher volume of calls into United's reservations centers. Our voice
recognition software was hampered by the volume, which in turn drove
longer wait times. We sincerely regret the inconvenience faced by our
guests, and are pleased to report that our voice recognition software
is fully up and running and wait times have been significantly
reduced. We are still experiencing higher than normal call volume,
which may result in sporadic call interruptions.


When the human model doesn't match the system model

"Sean W. Smith" <sws@cs.dartmouth.edu>
Fri, 25 Dec 2009 18:31:29 -0500

The real world gives another instance of what can go wrong:

A man in Epping tried to deposit $580 into an ATM but neglected to note the
transaction had not completed and walked away without the returned cash.
The next person pocketed the cash, but of course was identified.
[Source: AP item, 25 Dec 2009]

Sean W. Smith   sws@cs.dartmouth.edu  www.cs.dartmouth.edu/~sws/
Associate Professor, Department of Computer Science, Dartmouth


Disconnects between the Real World and Cyberspace

Bob Gezelter <gezelter@rlgsc.com>
Sat, 26 Dec 2009 08:19:04 -0500

Sometimes the real world and cyberspace are truly separate realms. The other
day, I experienced two episodes of just such a disconnect.

I was trying to verify the hours for a branch bank in the area. Checking the
bank's www site, I discovered that the branch was not listed. Calling the
customer service line, I was assured that if the branch was not listed, it
must have closed.

Three hours later, I visited the branch. It was quite active, with no signs
of closing or relocation. A discussion with the officers revealed that I was
not the first person to note the error. It had been reported to corporate,
but for some reason the data had not been corrected.

The same day, I had a similar experience with a major international
distributor. A local branch was not listed as one of their locations, even
though it was open and active.

It is an interesting question of IT governance when a company is unable to
keep its list of branches up to date.

A longer discussion of this episode can be found in "Bricks and Mortar
Hidden by Cyberspace".
http://www.rlgsc.com/blog/ruminations/bricks-and-mortar-hidden-by-cyberspace.html

Robert "Bob" Gezelter, +1 (718) 463 1079   35-20 167th Street, Suite 215
Flushing, New York  11358-1731 http://www.rlgsc.com


Obscure GPS problems not just in remote areas

Jeremy Epstein <jeremy.j.epstein@gmail.com>
Tue, 15 Dec 2009 16:02:12 -0500

The discussion of GPS issues with BMWs reminded me of a recent GPS
navigation problem I ran into using my brand-new Garmin nuvi 275T.

The Garmin maps of Washington DC have an unfortunate mistake: they don't
understand that K Street runs *under* (and parallel to) the Whitehurst
Expressway.  The Whitehurst Expressway intersects Key Bridge (well, actually
it runs into M Street a block from Key Bridge), but K Street does not except
when separated by 100 vertical feet.  Specifically, if you ask the GPS for
directions from northwest Washington DC (I was coming from the Tenleytown
neighborhood) to Fairfax VA, it tells you to take Rock Creek Parkway south,
exit onto K Street west, then drive along K Street and then make a left on
Key Bridge - which is 100 feet above you.  Google Maps, on the other hand,
gets this right.

The RISK is thinking that GPS errors only occur in out-of-the-way locations.
This is in the Georgetown neighborhood of Washington DC, hardly remote!

  [Alarmin' Garmin Swarmin' Harmin' not Charmin'.  PGN]


On the Road with a GPS System

Gene Wirchenko <genew@ocis.net>
Tue, 15 Dec 2009 18:46:58 -0800

  [Whenever I fail to run an item you think I may have missed, please
  resubmit (with "notsp" in the subject line, of course.  This is an
  instance of something I apparently missed 3.5 years ago.  Sorry!!!  PGN]

GPS problems have been in the news more lately, so I thought that I would
resubmit this item.  I have since moved back to Canada.  A key point: All of
the events that I describe happened on a trip to my mom's and back.  This is
not a collection of events over a period of, say, months.

On the Road with a GPS System

There have been previous submissions of the joys of GPS systems used in
driving.  They have been brief.  This writeup goes into much more detail and
details more than one risk.  I wrote this in June of 2006, but have kept the
present tense.

I am a Canadian citizen, but I live in Bellingham, WA, USA and work near it.
At the beginning of April, I went to visit my mother who was then living in
Greenwood, BC, Canada.  At the car rental place, because neither of the cars
available was acceptable -- perfumed cleaners are a non-computer risk some
face -- and because I am a frequent customer, I was upgraded at no extra
cost to a SUV, a SUV with a GPS navigator: the NeverLost system.  I did not
get lost, but that was not entirely due to the device.

I lost little time in starting to play with this new toy.

I found it interesting all of the streets that were nearby.  Unfortunately,
sometimes, the names were truncated, and in at least one case, the truncated
name made sense as a word.  I did not see any way to adjust the
magnification.

I saw that I could ask for a course to my mother's.  Unfortunately, the
interface was a bit confusing and rather than selecting the city of
Greenwood, BC (the smallest incorporated city in Canada), I accidentally
selected the then-current city and a street.  There is a street named
Greenwood in Bellingham.  There is also one in Lynden (near to Bellingham).
I found myself being directed to the south.  My mother's home was north and
east.  The device was quite persistent: "Please proceed to the designated
route."  Later, it got desperate, words to the effect of "When legal, please
make a U-turn."  I finally shut it off.

I tried again later, and it worked, mostly.  My mom lived on Kimberley
Street.  It is one of, if not the, longest street in Greenwood.  The device
did not know its name.  It did know many of the other streets, including the
cross-street by my mom's then-home.  The cross-street is two short blocks
long.  I finally programmed the device for the city centre of Greenwood.

I crossed the border at Sumas.  I then proceeded to Hope.  While on the way,
at one point I glanced at the display to see that a road was displayed to
the right.  This was disconcerting as that was where a cliff was.  The road
turned out to dead end.  I think it was a deactivated road, possibly the old
highway.

I usually stop at a certain gas station in Hope to buy a sandwich and drink.
In that area, there are under- and over-passes.  The device got confused.
It got its revenge shortly.  When I restarted the SUV, I looked at the
display and was confused.  The USA operates on the Imperial system of
measure, and Canada uses the metric system.  The device, recognising that it
was in Canada, had switched to metric.  When I looked more closely, I could
see the miles display.  It was rather smaller than the digits.

The gas station is on the old highway which parallels the current highway.
To get onto the current highway involves a few tight turns.  Some of these,
the device knew and some it did not.  I was directed to make turns when
there was no other road, but sometimes, the curve had no instructions.  This
also happened when I left Keremeos, BC.

The device warns about 2 Km before a turn and then just before.  The turnoff
for the Hope-Princeton Highway had already diverged from the main highway by
nearly one lane-width before the device announced the just before warning
that I should turn.  I was already in the correct lane.  Had I not been in
the correct lane, I could not have safely switched.

Not long after, I looked at the time display.  As it was just after noon, I
could not tell whether it was the time I would arrive in Greenwood or the
time left before I would arrive in Greenwood.  Later, I found that that
detail was documented on the card that was dangling from the unit, but many
other things were not.

While driving on the Hope-Princeton Highway, I found that many roads that
had no names were on the display, but that some roads that had names were
not present.

The Hope-Princeton Highway does run through wilderness, and in many places,
there are no other roads nearby.  That did not stop the device from telling
me three times each way to "Proceed to the designated route."  The voice is
a bit startling when it is not expected.

Glancing from time to time at the time-to-go display, I got suspicious.  It
seemed to be assuming 80 Km/h.  This is the default highway speed in BC, but
the Hope-Princeton Highway is known for being rather curvy in places.  It
has plenty of signs suggesting slower speeds.  At one place, the advisory is
20 Km/h.  In the summer, one can go bombing along much of the highway.  In
the winter, the curves get more dangerous, and you had better slow to 20
Km/h on the worst curve.  This was near the end of breakup, so conditions
could vary considerably.  What was the device assuming?

Princeton is at the west end of "The Regional District of
Okanagan-Similkameen".  The device displayed this from time to time clipped
to "Okanagan-Similkameen Region".  Much of this time, it displayed this next
to the road on the other side of the Similkameen River.  There was no
differentiation of region or city names from street names.  That road across
the river is called "Old Hedley Road".  When it finally was displayed, well,
the device leaves off the road designation.  The device did not display the
river though earlier, it had displayed much smaller creeks.

Leaving the village of Keremeos, BC, the road winds up a hill.  At the
bottom of the hill, there is a curve to the right (which the device told me
about), a curve to the left at the top of the hill (which the device did not
tell me about), and finally a right turn to the highway I wanted (which the
device told me about).  The up-shot was that I was told to make a right turn
followed by a right turn, and it did not make sense.  It was good that I
knew the road.  Had I blindly followed the advice, I would have gone down a
60 (or so) degree slope.

In the hamlet of Cawston, BC, the device thought that the main street
dead-ended and suggested accordingly.

I get paid in US dollars, so I prefer to spend US dollars.  Therefore,
rather than continuing through Canada, I cross at Nighthawk, WA, proceed to
Oroville, WA, fill up with gas, and cross back into Canada at Osoyoos, BC.

The device kept trying to get me to turn around.  I was wondering when it
would give up.  While I was wondering I saw a road displayed on the device,
but there was no road there, no sign of there ever having been a road there,
and no sign of anything else that the device should know about (such as a
creek).  A few miles past the border crossing, there is another road, a real
one.  The device seems to understand the world as segments.  When I got off
the segment, it recalculated.  The road to Oroville is a nice drive, very
pretty.  At least one of the roads that the device shows is actually a long
driveway.

Ah, Canada.  The device was displaying Imperial.  Remember how the device
switched measuring systems in Hope?  It switched again after I restarted the
SUV at the gas station in Oroville.  The system in use depends not on what
country you are in, but what country you were in when you started the
vehicle.  I found later that one can force the setting, but I did not
experiment to see if that locked down the measuring system used or that it
just lasted until the next vehicle startup.  Imagine explaining to a border
officer that you are experimenting with your GPS.  Guessing is such fun.

The device also had an odd idea of which way I should go.  As far as I can
tell, it intended to keep me on the highways.  I know of a shortcut, and I
took it.  The device did not handle it well.  Trying to get me back "on
course", it suggested some bad ideas.

1) One road to the left comes down a hill which is steep enough that the
road is broken into left and right branches.  The device suggested that I
take the far branch.  This would have necessitated a turn of approximately
150 degrees.  The near branch is about thirty degrees.

2) Later, it suggested that I take a right turn -- remember that the
previous suggestion was for a left turn? -- onto a narrow road with patched
potholes when I was on the best road around and which led straight to the
programmed route.

3) I rejoined the programmed route and prepared to turn right.  The device
was instructing me to turn left!

One thing that I never did solve was how to get the device to just display
without having a route programmed.  I wanted to see where I was without
having to select a destination.

The device has a safety feature that disables most of the user interface
when the vehicle is in motion.  It was also mounted so that the person in
the front passenger seat would not be able to see the display.  I thought
this was rather counterproductive.

When I set out to go home, I planned again to stop in Oroville for gas.
While I did not know the exact address, I thought I could get close.  For
some reason, the highway was not listed.  Unfortunately, the device requires
you to select first what you want (nearest city centre, a particular city
centre, or a particular intersection) then the city name.  The other way
around is much more natural to me.  It also would have been much quicker as
entering alphabetic was a slow process with no keyboard.

Again, I crossed the border at Nighthawk.  I crossed into the US a final
time.  The border between Canada and the USA is at 49 degrees north
latitude.  For some reason, the device told me I was in the USA when it
still displayed me at seven seconds of latitude north of the border.
According to my estimate, I was much closer.

I decided to let the device tell me how to get to Bellingham.  Understand
that I take a short route.  Roughly, I go west, then I go south.  The system
picked a route that was much longer and windier.  Among other places that I
saw, the oddly appropriate Chance Road.  In Osoyoos, it was trying to keep
me on the highways.  Here, it avoided them until the end.

I did make it home safely, and I certainly see where these devices are
useful, but much salt is needed.


GPS ads for captive bus riders (Re: Sullivan, RisKS-25.87)

<jidanni@jidanni.org>
Fri, 18 Dec 2009 10:57:52 +0800

Like the in-bus GPS "next bus stop" LED marquee boards here in Taiwan.
Between stops they have been programmed not to waste an idle moment, but
instead scroll "thank you for riding XYZ Bus Co." to riders
concentrating on them to learn the next stop's name.
At least accompanying audio just chants the stops.
Naturally everything must be repeated for each local language too.


Cruise control failed to disengage

Steve Cody <steve@codygang.net>
Sun, 20 Dec 2009 13:41:37 +1100

Similar to recent reports of uncommanded acceleration, we had an incident
where cruise control could not be disengaged.  This link should bring up
related news items..
http://news.google.com.au/news/more?um=1&cf=all&ned=au&cf=all&ncl=diIb-MX_41mrfpMRXwbgO6EIxST7M
Or google news for "Cruise control Frankston"


Re: LED Traffic Lights are efficient but cannot melt away snow

John Johnson <jvj@golden.net>
Wed, 16 Dec 2009 08:05:08 -0500

The problem is also evident on motor vehicles with LED signal and stop lights.
Snow is not melted away by the signal and stop lights and
accumulation blocks the lights.

new item:
http://www.newser.com/story/76251/led-traffic-lights-efficient-but-cant-melt-away-snow.html

Please report problems with the web pages to the maintainer

Top