Search within Lanny's blog:


Leave me comments so I know people are actually reading my blogs! Thanks!

Wednesday, January 28, 2009

AI and Robots: Insurgents "Hack" U.S. Drones

The Predator series of unmanned drones have been great weapons for the US government to fight terrorists and insurgents (see my previous posts). The US Department of Defense actually plans to replace one-third of military planes with unmanned drones. Any sensible person would have agreed that these unmanned drones are very sophisticated and advanced technologies highly classified because of the military and intelligence associations. However, on December 17, 2009, an article on Wall Street Journal reviewed that:
Militants in Iraq have used $26 off-the-shelf software to intercept live video feeds from U.S. Predator drones, potentially providing them with information they need to evade or monitor U.S. military operations.
WHAT???!!! When I first heard this on the NPR radio, I was on my way home, and I almost swerved off the road at this shocking news! This is unbelievable! I know the video downlink from our research UAVs are not encrypted, which means anyone with an antenna, a comm box, and a video capturing software could download video feeds from our planes. But come on! We are talking about 12 million dollars drones used by the U.S. Air Force and the CIA to fight real wars! Haven't they ever heard of the word "Encryption"? Even my neighbors' wireless networks are encrypted and cannot be accessed without a password. This is simply beyond my pithy understanding! The article used the word "hack" in the title. Did the insurgents really had to hack? The door was wide-open.

Imagine a bored terrorist pulling his laptop out to kill some free time by watching some real-time war clips, then he sees his terrorist friend's bunker in the live video feed, so he calls his friend and says, "Start your SkyGrabber program, man, I think that's your bunker!"
U.S. military personnel in Iraq discovered the problem late last year when they apprehended a Shiite militant whose laptop contained files of intercepted drone video feeds. In July, the U.S. military found pirated drone video feeds on other militant laptops, leading some officials to conclude that militant groups trained and funded by Iran were regularly intercepting feeds.





According to Dan Verton, a cyberterrorism expert, "we thought that this particular enemy was either incapable or not interested in learning how to do this...we've always been wrong on both accounts!" This is simply amazing! Didn't we know there's the Internet, and you can find tutorials for anything you want? Amazing!

"It is part of their kit now."


That's got to be the best line of the story! Especially at the thought that maybe they didn't even have to buy multiple copies and simply used pirated copies.

What's the merit of the story? Human stupidity is far more powerful than machine intelligence! As a matter of fact, I'll make that my Tao of the Day!





Human stupidity is far more powerful than machine intelligence!

Tuesday, January 27, 2009

Robot of the Day: MQ-9 Reaper

Since I am really in this UAV mood, let's talk about another UAV (Unmanned Aerial Vehicle) today.

The MQ-9 Predator B Reaper is a scaled up version of the MQ-1 Predator UAV we discussed in the previous post. It is also designed by General Atomics Aeronautical Systems, only larger, faster, more powerful, and much, much more deadly!

Picture credit: U.S. Air Force

Reaper was designed to be a hunter-killer, hence the name. It has a wingspan of 66 feet (20.12m), can fly at an impressive 300 miles per hour, and can stay in the air for up to 30 hours. The Reaper UAV is still powered by propellers at the rear end of the plane. It can carry 3800 lb of weapons. For example, it can carry 14 Hellfire missiles or other weapons such as the 500-pound, laser-guided bombs shown in the picture above. The U.S. Air Force was activated to operate the MQ-9 Reaper UAVs on May 1, 2007.

Fully loaded Reaper!





Looks like the U.S. Customs and Border Protection also upgraded their UAVs to the Predator B Reapers!




Picture of the Day:

Disclaimer: This is not my house. Read the story below!

 

"Good news is that I truly out did myself this year with my Christmas decorations. The bad news is that I had to take him down after two days. I had more people come screaming up to my house than ever. Great stories. But two things made me take it down.


First, the cops advised me that it would cause traffic accidents as they almost wrecked when they drove by.

Second, a 55 year old lady grabbed the 75 pound ladder almost killed herself putting it against my house and didn’t realize that it was fake until she climbed to the top (she was not happy). By the way, she was one of the many people who attempted to do that. My yard couldn’t take it either. I have more than a few tire tracks where people literally drove up my yard."

Monday, January 26, 2009

Robot of the Day: MQ-1 Predator

The MQ-1 Predator is probably the most famous UAV (Unmanned Aerial Vehicle) of all times. It was developed by General Atomics Aeronautical Systems for the USAF (U.S. Air Force) and the CIA.

Photo credit: U.S. Air Force

The MQ-1 Predator UAV has a wingspan of 47.8 ft (14.8m), can fly a maximum of 135 miles per hour, and can stay in the air for 14 hours. The cost for an early production was around $3.2 million.

Initially it was only a reconnaissance system allowing the remote operators to acquire aerial video in real-time. After the CIA deployed the Predator UAVs to Afghanistan, they expressed strong desire to add the capability of firing Hellfire missiles from Predator UAVs to kill terrorists. So it was done. On February 4, 2002, a CIA Predator attacked a convoy of sports utility vehicles, killing a suspected al Qaeda leader who the CIA thought were Osama Bin Laden.

The Predator UAV requires a satellite link and is operated by two pilots (most likely in a military base in Nevada) sitting in front of cockpit like devices. The control of the UAV falls under the tele-operation category because most decisions are made by human operators.

The first video below showcases the capabilities of the MQ-1 Predator to quickly track down a moving vehicle (note that it is much easier to tracking a lone moving car in a desert compared to tracking down the same car, say, in LA traffic). The second video shows firing of the missiles.






An unknown number of Predator UAVs are also used by the U.S. Customs and Border Patrol. I would guess these Predators don't shoot missiles at illegal aliens.

Picture of the Day:


Residents in Norway were stunned by the beautiful yet mysterious light show. Turned out it was caused by the malfunction of a Russian missile test. Follow this link to read more.

Saturday, January 24, 2009

Paper Review: A Comparison of Event Models for Naive Bayes Text Classi cation

This paper is written by Andrew McCallum (Just Research) and Kamal Nigam (Carnegie Mellon University) and published at an AAAI-98 workshop on learning for text categorization. This is a seminal paper cited by 1426 papers according to Google Scholar!

Two text classification approaches, the multi-variate Bernoulli model and the multinomial model, both use the naïve Bayes assumption. This paper tries to differentiate the two models and compare their performances empirically on five text corpora.



In text classification, a naïve Bayes classifier assumes that the probability of each word occurring in a document is independent of the occurrence of other words in the document and is independent of the word’s context and position in the document. This assumption simplifies learning dramatically when the number of attributes (features) is large. Both approaches use training data to calculate estimates of the generative model and then uses Bayes’ rule to calculate the posterior probability of each class given the evidence of the test document. Then the class most probable is selected.

One major difference between these two approaches is that the multi-variate Bernoulli model doesn’t care about word frequency in documents, while the multinomial model does. Another difference is that the multi-variate Bernoulli model explicitly includes the non-occurrence probability words that do not appear in the document.

When selecting features, in order to reduce vocabulary size, only words that have the highest average mutual information with the class variable are kept. Average mutual information is the difference between the entropy of the class variable and the entropy of the class variable conditioned on the absence or presence of the word.

Five text corpora are used and they are Yahoo! ‘Science’ hierarchy, Industry Sector, Newsgroups, WebKB, and Reuters. Empirical results show that the multi-variate Bernoulli model works well with small vocabulary sizes, but the multinomial model performs better at larger vocabulary sizes. The multinomial model produced on average a 27% reduction in error over the multi-variate Bernoulli model at any vocabulary size.




The boringness of a paper is inverse proportional to the amount of time it takes to put the reader to sleep.




Video of the Day:


You have to watch past 0:40 to really appreciate it! It's an LCD!

Friday, January 23, 2009

Random Thoughts: Y2K all over again in 2010

Many of you probably still have some vague memories about the Y2K bug. Come on, it was only 10 years ago. Many people thought it was going to be the end of the world for all computers; catastrophes caused by failing computers were going to happen everywhere; people were going to be killed!


So what happened? Nothing major really? A little glitch here and there, but people kept living their normal life. Why such a big hype before, but so quite afterward? It was because many programmers and engineers paid great attention to the Y2K bug and worked diligently before the millennium passing to make sure things went okay. I could clearly remember where I celebrated the new millennium at -- next to my servers at work, to make sure nothing fails when time went pass 0:00am on 1/1/00.

So what was really the Y2K bug. In simple words, a lot of programs only use two digits to store year and assumes the first two digits of the year will always be 19. that means after the last two digits turned from 99 to 00, the first two digits remained the same, resulting in 1900.

Interestingly enough, when everyone's clock went from 11:59pm 12/31/2009 to 12:00am 1/1/2010, a bug similar to the Y2K bug hit the world (especially Europe) and caused all kinds of havoc:
  • ATMs and point of sale machines rejected debit cards of 30 million people in Germany since New Year's Day
  • Similar problem occurred in Australia where point of sale machines skipped ahead to 2016 rather than 2010
  • Symantec's software (anti-virus, anti-spam, etc.) treated all new updates from the company as old data and refused to update
  • Some users of mobile phones also reported that they started to receive messages dated in the future: 2016
  • Palm had to release a new version of the operating system so their customers' Palm Pre phones would continue to sync and the calendar program would continue to function
I feel really sorry for the many Germans who got stranded at ski resorts because they had no way of paying hotel and restaurant bills. Many people also had to stay in long lines to withdraw cash from the counter. This really goes against the holiday spirit! But who is here to blame for this Y2010 crisis?

A French (credit) card manufacture, Gemaltao, had admitted that the problem was because of a programming failure, which affected chips on credit cards and debit cards. The estimated damage is over €300m.

So how could a software bug cause 2010 to turn into 2016? My guess would be that the programmer used a hexadecimal data type instead of the decimal data type for the last two digits of the year field. Let me explain here.

For decimal numbers, each additional digit would mean 10 units of the previous digit. For example:

10 = 1 x 10 + 0
15 = 1 x 10 + 5
123 = 1 x 10 x 10 + 2 x 10 + 3

For hexadecimal numbers, each additional digit would mean 16 units of the previous digit. Therefore, the same representations in the previous examples would mean very different numbers in decimal:

Hex 10 = Decimal 1 x 16 + 0 = Decimal 16
Hex 15 = Decimal 1 x 16 + 5 = Decimal 21
Hex 123 = Decimal 1 x 16 x 16 + 2 x 16 + 3 = Decimal 291

If you look at the first example closely, you'll see that year 10 in hexadecimal turned into year 16 in decimal. This also means the program at fault still only used two digits to represent the year field (probably assuming 20 for the first two digits to save memory), but used the wrong data type.

The other merit of the story is that when people are relying on technology more and more nowadays, what consequences do we face if some key technologies fail? We can only do so much to anticipate certain failures (such as the Y2K bug), and when failure strikes, we suffer.

To think further, what if the failure were caused intentionally by criminals? I don't see them taking over the world easily, but they can certainly cause a lot of damage (disruption of power grid comes to mind). How can we get ready to deal with this kind of challenges? I guess that is an open question still waiting to be answered...


"Always code as if the person who ends up maintaining your code will be a violent psychopath who knows where you live." -- Martin Golding





Picture of the Day:


Look at all those people stranded at the Austrian ski resort!! :)