Hacker News Reader

Comment by d--b | Hacker News Reader

Unless you need a captcha in an airport, I wouldn't say this qualifies as useful... 0.7% error rate on biometrics means that of the 1000 people who will use the airport on a particular day, the system will fail 70 times, that's huge. Plus this is in lab conditions. Add a few years to the sensors, and your error rate goes through the roof.

Plus I'm pretty sure you could develop "adversorial shoes" for this...

Meh.

View on HN

What do you think the current error rate is? In testing the TSA has something like a 100% failure rate iirc.

I'm no fan of the TSA, but you're comparing apples to handguns-in-handbags here; their inability to find planted explosives in my luggage isn't the same as a failure to correctly identify that I have the right documentation to get on the plane.

You're right; failing to detect weapons (and thus letting people die) is a fatal error that can't be corrected. Failing to correctly id you is something that can be corrected by a human afterwards. I'm much happier to let a system incorrectly id someone than to let them onto a plane with a weapon.

That said, this seems like just one of many potential id methods that would comprise a larger system (facial recognition, fingerprinting, etc).

ORD had 79,828,183 passengers in 2017. Even if you cut the error rate in half you're still looking at an error once a minute every hour of every day.

OHD has 41k badged staff. Enough staff to triage 2 situations a minute constantly would barely make a blip.

It'll fail .7% of the time for each read. You must assume you will be read multiple times at multiple locations under different lighting and motion tracked the rest of the time. Once you get a couple consistent identifications, the system can be pretty sure it's you.

>Plus I'm pretty sure you could develop "adversorial shoes" for this...

Add/remove some weight from your non-wheeled luggage.

Reply on news.ycombinator.com