Hacker News new | ask | show | jobs
by wraithm112 3836 days ago
The problem with this theory, as I see it, is identity. When you go through the body scan, they don't ask for your identification. Yes, they ask for ID before going to the security line, but there's a very difficult combinatorial problem with lining up IDs with bodyscans.

Also, I'm inclined to believe that the TSA is incompetent all the way up.

3 comments

You don't actually need to match up to the individual, you have roughly say 100 or 1000 possible candidates for each scan, based on people you processed in the last X minutes. You can just save that then all you need is 2 scans to be able to match.

If you're tagging face when they get processed and running face recognition at the scanner, (which is either already possible or will be possible within a couple years) you don't even need to do that.

On a related note, I feel like people dont factor in the (exponential) progress of technology in their threat assessment. For example sites that tell people their password strength don't count Moore's law into password difficulty calculations. Nobody uses post quantum crypto. The NSA has played this by storing as much interesting encrypted data and then decrypting it when technology progresses Or vulnerabilities are discovered. Most of our encrypted data is plain text in some X number of years.

Anyway going back to scanners, you could just store data and wait until computer vision gets to the point where you can do it with computers. Anything a person can do a computer will eventually be able to do. If a person could watch a bunch of camera feeds and track each person from the point he presents ID to the point he is scanned a computer will eventually be able to too.

IMO they already can.

Except they don't know who they've processed. I guess they have the list of ticketed passengers, and most people will only go through the machines once. But they don't record anything when they check your ID.
I don't think this is a very hard combinatorial problem. At least last time I flew, they scan my boarding pass when I'm pretty close to the scanners. They'd just have to pick me out of, say, 50-100 other scans. And once they establish decent time data on BP scan to body scan correlation, they'd be able to narrow the field further.

I agree with you, though, that your last point makes this moot. They're too bumbling for me to worry.

This objection seems... naive? Taking an engineering mindset, matching up people and scans doesn't seem like a very hard problem. The checkpoint areas are blanketed in cameras, so face recognition is one possibility. Or, an easier solution would be a computer vision system that tracks movement.
If they have cameras that can track people via facial recognition, then why would they need to track people via body scans?

If they really wanted to use body scans to track people, they'd arrange the checkpoints so the body scanner is the first place you go after they scan your boarding pass and/or ID. And when questioned about why they rearranged the checkpoints that way they'd just say "TSA security evolves to counter ever changing security threats, and we can't speak specifically about what prompted this reconfiguration." End of story since security theater is not to be questioned.

Facial recognition is fragile - a simple prosthetic or three can throw it off. Body recognition (and gait recognition) is less so, mainly because there's relatively MORE of it that would need to be disguised.