Identifying a person through walls from video footage, using only WiFi
techxplore.com
techxplore.com
>The lab has tested their new technology on 1,488 WiFi-video pairs, drawn from a pool of eight people, and in three different behind-wall areas, and achieved an overall accuracy of 84% in correctly identifying the person behind the wall.
What does it mean exactly? They only need one short video and one short wifi capture to get 84% success rate? That's what the video seems to imply but I find that very hard to believe. Or maybe it's just because it's fairly easy to distinguish among 8 people (especially if they have significantly different body types) and it won't work quite as well at large. I can identify my girlfriend's footsteps in the staircase with remarkable accuracy but I can guarantee you that it won't scale to the general population.
Maybe it works better than I give it credit for but they need to bring up better evidence IMO.
My understanding is that the video on the left (the one showing the other side of the wall) is just here to illustrate what's the wifi setup senses and was not used for identification as the whole point is to identify without (optically) seeing what's on the other side.
Still, that leaves the problem of the "pool of eight people" and how this technology would work in less controlled environments.
Key aspects upon this are that it works by matching up their style of walking (gate), so the person needs to be moving not only in the wifi data capture, but also the video they are matching with.
They use the video to extrapolate a 3D mesh model of that person walking. Then model a wifi signal upon that model and produce a signature for that person that they can then use to match up with what the wifi signal is picking up.
Whilst measuring some bodies way of walking has been done many times before, this add's another way of doing that - but without the need for a camera feed.
So with this approach - no they can't produce a video image of the scene behind the wall from the wifi signal(s).
Remember - individuals have many unique traits, many of which we havn't tapped into yet, though as far as walking styles go - that has been done for years, this just creates another way of measuring those and matching up with other forms of data, video in this case. Other things that could be explioted this way would be for example sound - a contact mic array would allow you to (thru a wall) measure a person's heartbeat to some degree and match that up to an individual. Privacy wise, this may well traction well compared to other forms of surveillance.
Goodbye 4th Amendment.
Would be cool if these intelligent folks but their efforts towards improving privacy.
This is a neat application of WiFi signals, but the accuracy is nowhere near good enough for this to reliably ID random people.
I’m way more worried about online corporate surveillance that actually can ID me reliably with 100% accuracy and get a lot more info about me besides where I am, than I am of government surveillance of my walking gait through walls.
Only one of these is objectionable.
The other two ought to be pursued so we can further advance the human race and improve the quality of everyone's lives. They're totally justified.
The thing is though, money does that already if your society doesn't have strong inheritance taxation, so I'm not sure it's really such a problem.
That's not a threat to humanity.
> And Costa Rica?
> And most individual countries in Europe?
All of these countries, for practical purposes, do not exist.
They do not have influence on the world stage. They may as well be China, considering how easy these countries will bend to Chinese money.
So either you develop the surveillance tech, or you must look the other way when China develops it. If you disagree, your economy will be ruined.
The whole developement of nuclear weapons is a story of the when, of the how fast and of the should I and would have looked certainly different if it’s protagonists all had based their decisions on the ethical rationalisation you proposed here. Who knows, maybe even the Germans would have gotten nuclear weapons?
The hard question obviously is: where do you draw the border? If you invented the paperclip and it was beeing used in the holocaust’s bureaucracy blaming you would be a bit far fetched. If you invented the gas showers of the KZs whose only purpose it was to mass-kill prisoners, not feeling ethically responsible would be totally mad. In between there you have got Werner von Braun with his V2 rocket, which isn’t exactly easy to judge. He enabled space travel, yet the bombs killed many british civilians quasi randomly and out of the blue and the number of slave workers killed during construction was even higher than the number the rocket killed.
Inbetween these three examples there are infinite shades of ethical dilemma while your argument paints this as a binary question about whether something could or should exist or not — and if the answer to the could is True, who cares about the answer to the should, right?
I propose something else. If you value your own intellect you should abstain from ideas that could be, but shouldn’t and rather attack things that could be and should be or maybe even things that should be, but couldn’t by the thinking of today.
We engineers, scientists, thinkers and artists are giving birth to things that become reality, but it is our indivisual decision if we do:
You will rarely find a writer telling you he had to write that dehumanizing dangerous text, because if he wouldn’t have done it somebody else would have. We can stray farther from this even and paint an example where a old lady collapses at a bus stop. Now you could have someone robbing her as she lays there (”If I wouldn’t have done it someone else would’ve”). Or you could have somebody going for first aid and calling the ambulance (”If I wouldn’t have done it someone else would’ve”).
The statement “someone else would’ve done it” is not really a ethical reason to do anything. The question you should ask yourself is: in what kind of society do I want to live and do my actions contribute to it and if so at what cost? What if everybody acted like I did?
But then of course this is the wrong question to ask. This is nothing more but a complicated version of a multi-party, repeated Prisoner's Dilemma problem. Except that you don't know your negative payoff when an opponent chooses to defect, but it may well be minus infinity (you get holocausted or nuked into paleolith). And then defection by any single party gets you the negative payoff. So if you only ask the question you suggest and ignore the rest of the options, you are doomed.
I'm afraid that the world's politicians (most of them at least) are completely sane, and all this "race to the bottom" stuff going on is just the result of the game theory & life being what it is.
I think context is key, and we shouldn’t theorize conflicts in everyday life in the same way as all out nuclear war or other existential situations. Maybe that is just the naive idea of a guy who benefits from the wealth and peace produced from a Europe which for the first time in centuries doean’t feel the need to kill each other for (in hindsight) more or less meaningless reasons.
Bad things still happen even they’re outlawed, but that doesn’t mean we just throw up our hands and proclaim that there’s nothing anyone can do about anything, and dismantle the entire legal system.
Someone will do it, but someone will also do the counter-thing to it. In this example someone might research how to prevent people from being identified through walls, and I would hope that those people are the more skilled ones.
As tech marches on, "not commonly available to the public" becomes a smaller space.
Separately, I'd hazard a guess that they still use FLIR in an indiscriminate way, and just conjure up a little parallel construction when needed.
I share your concern for what will become of privacy as exotic new surveillance techniques become widespread. The legal treatment of expectation of privacy seems tied to whatever is the societal norm of the age, and that mark has shifted dramatically in my lifetime.
There's way more money in bootlicking
plenty of intelligent folks working on this for example
So publicly sharing this research is an effort towards improving privacy, by letting other researchers & the public know what the risks are.
The thing about a lot of machine learning tools is they can get better and better over a period of a decade and yet still remain undeployable - see self-driving cars.
I mean, the hypothetical suspect situation would like be extremely messy and getting a clear scan to deploy this seems implausible. General purpose surveillance would probably just use cameras intended for the purpose.
In all seriousness, the confidence interval of this is really low - 83% among a tiny pool of individuals. It's really interesting to be sure, but the tiny amount of confidence would make it extremely unlikely to be used in the context of justice/law enforcement (then again, maybe not [2]).
[1] https://www.youtube.com/watch?v=78OKaiYld1E
[2] 10% false positive rate in active use: https://www.smh.com.au/national/nsw/mobile-cannabis-tests-on...
"The lab has tested their new technology on 1,488 WiFi-video pairs, drawn from a pool of eight people, and in three different behind-wall areas, and achieved an overall accuracy of 84% in correctly identifying the person behind the wall."
Yeah I often do exactly 62 trials with 24 combinations (eyeroll)
Of course, even if all people involved were white, and even if it was a deliberate reference, as long as there's nothing other than the number, no one should get in trouble for that. Under no circumstances should researchers be worrying about the numbers they come up with. If you're doing 420 trials at 69 degrees Fahrenheit, so be it.
Cf. this thread: "Six paediatric health‐care professionals were recruited to swallow a Lego head. Previous gastrointestinal surgery, inability to ingest foreign objects and aversion to searching through faecal matter were all exclusion criteria. Pre‐ingestion bowel habit was standardised by the Stool Hardness and Transit (SHAT) score. Participants ingested a Lego head, and the time taken for the object to be found in the participants stool was recorded. The primary outcome was the Found and Retrieved Time (FART) score." https://news.ycombinator.com/item?id=18519899
While I do believe something like this is possible, the evidence of this actually working is not really convincing.
They use the 3D mesh to model how they think they signal would be modified by the person behind the wall, then they compare the received signal with their model to see if it's a match.
> Can a pair of WiFi transceivers outside the house determine if the person inside the house is the same as the one in the robbery video? Questions such as this have motivated this new technology
We have windows, tiny little cameras that go through walls, and we can ping cell-phones. Correct me if I'm wrong, but wouldn't all of those work better than this incredibly convoluted method?
On a legal basis, however, in some states it's illegal to even look over a privacy fence without a search warrant. I imagine this approach is less accounted for by lawmakers of the past.
And since it could use your existing home Wifi as the source signal thus the sensor would be receive-only, it could be very battery efficient so could last a long time on battery.
This is research. Currently it's unreliable (83% is super low) and convoluted, but with more research and investments, it is possible that they would be able to make it reliable and practical.
The effect would be that you can't observe limb or limb part movement inside the skirt, but only how they move and deform the skirt volume. I think that's still enough, given that gait recognition works on people in burkas or other well covering garments.