KDS

I fully acknowledge what Dean reported for the reduced number of secondary / resolution checks as a result of his interactions. That can make the security process easier for those people.

My objection is the reaction of most TSOs when I choose to not participate in the discussion because I have no need for it (I always opt out). See this post for my reasons about this:
http://www.flyertalk.com/forum/check...l#post17813207

Dean likely moves on to next person if the passenger does not engage with him. Most do not...they insist on being acknowledged by the passenger -- like the used car salesman who won't leave you alone.

PTravel

HSVTSO Dean, thanks for an informative and interesting thread. "You da Man!"

ND Sol

+1 ^

HSVTSO Dean

Neither. I left it in the van out in the parking lot. :P

The second. Even before ATR, the machine didn't produce an "image" like a camera does. Everything about it was just raw data, ones and zeroes and such, and the programming of the machine turned that raw data into what we would call an image. From what I understand, installing the ATR basically goes something like a full software overhaul of the machine, and even the capability of producing that image -- this mangled mess here -- is gone.

What you see -- here -- is all that's visible on the screen. It's a touch-screen, lacking any kind of other input devices such as a keyboard or a mouse. The only options on-screen that you have are male scan (the blue one), female scan (the pink one), and clear, which purges the current scan to move on to the next one. There are also safe-guards built into the programing that prevents the operator from continuously scanning people and not clearing the avatar and data.

When I gave the machines a once-over, I didn't see any means for them to store or transmit the information. Like I said before, it's possible that such a thing could be internal to the machine, but I didn't see any direct connection ports on the machine itself (such as serial ports or USB slots or a CD bay. Just as an aside, the carry-on x-ray machines at the security checkpoint have all three). Short of tearing the machine apart piece by piece and deconstructing every line of code that went into its programming, I can't really give a 100% empirically-true answer to this question. But, just from what I saw, I'll say no.

The closest it came to it was what I think are scan numbers, which the supervisor occasionally rolled by and jotted down. In the image above - this one - it's in the very bottom-left corner. That particular machine, at the time that particular photo was snapped, had performed 385,607 scans (if that number is what I believe it to be). Supporting my educated guess on the matter is the fact that the same thing is done with WTMD devices, using a built-in counter on the machine. The numbers are written down hourly to track throughput. It's hardly the best picture for it, but you can (kind of) see the console up at the top of the WTMD. Those are buttons beneath the green/red displays.

One of the important things to always keep in mind when doing this job -- indeed, any job -- is to not take things personally.

Yup. Can usually tell right off the bat if someone is going to engage or not. If they respond to "Good morning, sir!" with "..." then it's a pretty good bet that they're not.

chollie

If the full-body frisk (required for those of us physically incapable of using the NoS or those who opt-out) only takes a couple minutes (supposedly), why do those pax have to wait for special treatment?

If I understand you correctly, two people staff the NoS, one of each gender, and you swap off as needed to resolve anomalies. Why not just 'feed' someone who requires a full-body frisk into this line, instead of pulling them out, often taking them out of sight of their belongings, requiring sometimes 3 TSOs - a 'caller/minder', a frisker and a bag searcher.

I know some folks don't want to walk through the NoS because they don't trust that the machine is truly off, but I think most folks (including me) would walk through it to get to the other side where able-bodied folks get their frisks, full or partial, as needed.

janetdoe

1. Every digital image is just ones and zeros; the machine was acting exactly like your digital camera, except the "flash" was microwaves instead of light.

2. From the work I have done with image processing software, I can guarantee you that the same ones and zeros are still collected and an image could be easily displayed out of this data. If my computer doesn't have a program that opens .tif files, that doesn't mean I don't have any .tif files on my computer. It may be a semantic distinction... I know your point is that the software would need to be modified for a TSO to be able to see the real image on the screen.

ATR is most likely a software update where they added some detection algorithms (for example, edge detection algorithms would detect where the body starts; the same algorithms could also detect sharp lines/gradients within the area of the body) and some processing software to draw a box on a stick figure in the approximate area of anything that was detected by the algorithms.

3. Thanks for the interesting thread.

Wally Bird

This aspect has been discussed before with dissenters claiming that the sanitized software would only detect sharp, straight(ish) lines and harsh contrast areas. The proponents, including the TSA, have not disproven or even denied this contention.

So do they work ? 100% no. 50% perhaps. Somewhere in between probably.

Xyzzy

Thanks fr the interesting writeup!

janetdoe

I missed those discussions, but at our lab, we have a simple $2000 software package called SPIP for analyzing images taken from microscopes (TEM, SEM, AFM, optical, etc.) and I assure you it can detect much more that straight sharp lines. A very simple example shows that it can distinguish a very noisy background from areas of interest.

Further, I was able to combine the image analysis software with some very simple neural net training, and the computer ended up being able to do classifications instantly and with higher accuracy than human classification. We would run the neural net algortihm on an image we had classified by hand, go back and see why the computer "missed" it, and more often than not, the human classification would be wrong.

And neither of these is my area of research or my specialty, just hacking around with some basic off-the-shelf tools. Based on what I was able to accomplish in a few weeks of work, I have no doubt that the image recognition and anomaly detection could be incredibly effective, especially if they tune the algorithms to accept a high number of false positives (which is certainly the case). Think about it, how advanced is Google image recognition software? And that covers ALL images, not just images of humans in a single pose.

Of course the 'pancake' problem could still be an issue, but frankly, that's a very theoretical example, and assumes that someone can find an explosive that matches the MMW signature of human skin, and then carefully and precisely sculpts it. Even then, if it could be done, it would not be hard for the ATR developers to take some MMV images of the human+pancake and identify the image or signal artifacts that could be detected using a different algorithm. <shrug>

The TSA would be incredibly stupid to confirm or deny the capabilities of the system.

mikeef

Sad to say, that was exactly my first thought.

Mike

janetdoe

Really? I wondered if the Librarian was hot.

mikeef

Librarians are always hot.

Mike

Wally Bird

I agree that publicizing the limitations would be counter-productive but we have empirical evidence of things getting past the scanners undetected.

That's either a failure of the machine/software or an oversight by the TSA operator(s) or both. I don't have a lot of faith in anything the TSA does (you may have gathered ) so I can't necessarily accept that the 'new' software is as capable as that you describe. It ought to be - it took long enough.

janetdoe

Agreed 100%. Just saying that technologically, what they have claimed is not unbelievable, and a 50% miss rate sound ludicrous even for an amateur programmer like myself.

I'll see your "I don't have faith in the TSA" and raise you one. The NSA could easily be reading the EM signature from the AIT system, capturing the images that way, and storing them in databases for analysis with facial recognition software. Let's see if we can figure out if MMW or BSX machines are TEMPEST-certified.

halls120

Well said. I refuse to be treated as a suspected terrorist simply because I decide to fly on an airplane.