Nick Vivion asks, “Could this new Wi-Fi technology revolutionize airport security?”
Some of the most promising new technology has emerged from a multi-year project from MIT’s Computer Science and Artificial Intelligence Lab: a Wi-Fi network that can identify who you are — even through a wall. Yep, you read that right. These geniuses have built a way to implement Wi-Fi as a means to identify the unique characteristics of individual humans.
The RF Capture technology is able to analyze how Wi-Fi signals bounce off a human being to create an outline similar to what you might see from a millimeter wave scanner. The secret sauce is a reconstruction algorithm that stitches the many refracted waves into an image and then analyses the results. The system was able to identify 15 different people with a 90% accuracy.
The last sentence is meaningless. There are two types of errors. False positives and false negatives. Which type is this 90% numbers? Furthermore there are two types meanings of “identify” in biometrics. There is, “Who is this person?” (identification). And there is, “Is this person who they say they are? (verification)”. If they are identifying one person out of a population of 15 90% of the time then the success with of a population of 100s, 1000s, or millions is going to be insignificant. If they talking about verification then it means that one out of ten times an imposter is falsely verified. This is way too poor.
If that isn’t enough they don’t address the fundamental difficulty with security. That is that you have an active adversary. The adversary is going to do whatever they can to fool you. Wi-Fi signals bounce off of humans, as shown in the video below, but they bounce of metal even better. A little bit of aluminum foil underneath your shirt and you will appear as a completely different image to their technology. Some outdoor clothing has aluminum built into it for heat retention. This would play havoc with their tech.
Vivion should have asked a security expert his question. The answer would have been “No. Next question.”
“It doesn’t work”, in Left -Speak, translates as “More funding is obviously required. Q.E.D.” So you have just told the trough-feeders to double down.
Reminds me of “smart gun” technology which I refer to as dumb liberal gun technology. They think that attaching the fancy tech word of Wi-Fi that the low information voters will think it is magical. It is just as flawed from the get-go as smart guns and I suspect will likely never reach the required level of sensitivity and specificity.
I think a lot of this newfangled tech they constantly talk about is garbage….but those with an agenda want society at large to think that capabilities exist that in reality are worthless. If the public at large thinks you can do a thing that’s damn near as good as actually being able to do it for real.
This illustrates how for the last fifty years people have gone into “journalism” because they aren’t smart enough to even make it through the degree program for an English Major.
This is “not ready for prime time” until it can identify and verify with six sigma accuracy. Otherwise it is worse than the “no fly” list that had Senator Ted Kennedy on it. Although keeping that creep on the ground was a public service while it lasted (Now of course, he’s safely IN the ground).
But yes, if people think the government can detect people this way and detect cash because of the plastic stripes in the bills, it is almost as good as if the government really can do those things.
FYI: the late senator kennedy (deliberately not capitalized) is deep underground and is still drunk in Hell where he currently resides.
Adversary?
Yes. Fixed. Thanks!
The tech is interesting but the overblown claims are BS.
Basically, they reinvented radar. Then added some processing to merge multiple scan into a single image. That’s mildly clever, but not new.
But the freq range they use (5.46-7.24 GHz = 4.1cm-5.5cm; that isn’t even wifi) cannot give them adequate resolution to distinguish anything but the largest, grossest features; forget facial recognition. They _might_ be able to couple it with experimental _gait_ recognition to ID someone they’ve previously IDed by other means.
“The system was able to identify 15 different people with a 90% accuracy.”
Total crap claim. They did _not_ identify anyone. They could _distinguish between_ people; at a 97.5% ‘accuracy rate’ when only trying to pick five people. when they increased to 15 people the rate dropped to 88.2%. The paper doesn’t specify false positives or negatives.
As currently designed, this probably wouldn’t work well in public. At the power level they specify, anyone walking by with a 802.11a/h/j/n/ac device in service would jam them.
The two leading authors on the paper are students. This appears to be a project to familiarize them with tracking and imaging principles.