2016: https://www.schneier.com/blog/archives/2016/08/keystroke_rec...

It's not much of a stretch to imagine that intelligence agencies have been heavily invested in this area and are far ahead of public research, given signals intelligence has basically been their bread and butter since forever. Moreover, Stuxnet was so advanced for the time that its existence stunned the world.

Keystrokes can be captured indirectly via audio analysis, electromagnetic emissions from wiring, and now RF imaging techniques looking at finger movements. Wouldn't be surprised if they can create multi-modal composite models to attain higher accuracy, or if RF imaging is able to capture lip/jaw movements these days.

The really sexy part is probably what they're able to do with fixed wing airborne platforms, where you can afford to pack ridiculously high-end sensors and local computing power on board.

It still weirds me out to think that a gimmick from 2008's The Dark Knight is more or less a reality now, or will be soon if it already isn't.

My first job in tech as a teenager (2007 or so) was doing blueprint analysis and WiFi Access Point placement for a US Defense contractor in the Midwest.

While working on one of the buildings with some missile guidance programs, I found a small room in the center of the building that had twelve inch thick concrete walls and a thick steel door. Determined to do my job, I experimented with placing several access points near this room until I found a combination that would force enough signal to connect through those walls. I had the telecom team pull wires, a month later I threw some WAPs in my backpack and installed them.

A week later I got an email marked urgent demanding that my team turn off these access points immediately. I complied, but asked what exactly the concern was. They mentioned that by bouncing WiFi signals, a van parked in the parking lot could monitor the activity in any room they wanted.

At the time I thought they were crazy, and at times I've told this story to demonstrate how paranoid that company was. Looks like there was some real basis to their concern.

BTW you don't need rocket science to find out who is back home in your neighborhood, or the other room etc. As most people carry phones in pockets/bags so it is enough to just log clients' MAC addresses in WiFi promiscuous mode and then correlate them.

Presumably, the techniques for lens-less cameras will still apply in the microwave spectrum - https://arxiv.org/pdf/1710.02134.pdf

By moving a wifi adapter in a 2d scan pattern, you could presumably create a virtual 2d sensor and then treat anything between you and where you are wanting to image as the diffuser.

This is new, the arxiv page says "Submitted on 23 Oct 2018".

(There has been previous research published from MIT on the same topic[1], so this was not ovious)

[1] https://www.technologyreview.com/s/415539/wireless-network-m...

I was curious how they measure WiFi signals in Android. It seems surprisingly easy with:

- https://developer.android.com/reference/android/net/wifi/Wif...

and

https://developer.android.com/reference/android/net/wifi/Sca...

which seems to give very fine grained information about WiFi strength of any network in the vicinity.

Quite an awesome usecase would be to use this for perimeter security, like securing your house and cars. In my locality camera's are in a legal grey zone: the police applauds them for use in case of crime, but because of privacy laws you're not supposed to film other peoples houses and whereabouts. This would enable you to track movements without camera's. Software would make it easy to only give you signals when somebody or something moves on your property at night. All I would need is some machine learning to learn about rabbits, cats and foxes. Or you could have cameras that only turn on in case of movement on your property, hence triggering (I would hope) the legitimate interest provision in the GDPR.

A whole article about seeing through walls, without a single photo?

This article implies that WiFi is a privacy concern, but wouldn't any other RF signal work just as well? A bad actor could just create their own signal if you somehow protect your WiFi. That would probably be more effective anyway because they would know exactly where the signal was coming from and could choose a frequency for this application.

HF motion sensors are now used in a lot of lights. They are 5.8 Ghz radars that detect motion even through walls.

So high frequency motion detection is already used in a wide range of applications.

But I think 'seeing' should be taken with a grain of salt. Yes you can detect motion behind a wall but creating an image is some steps away.

Cue military applications, where snipers can now kill you in the safety of your own home with a wall piercing bullet and a scope with a wifi based image overlay.

2015: https://www.youtube.com/watch?v=fGZzNZnYIHo

Time to put the wifi router on my kids' train set.

In https://news.ycombinator.com/reply?id=18398475&goto=item%3Fi..., foreigner quite reasonably asks, "Won't any RF signal work just as well?"

Any RF signal will work, but some work better than others. Wi-Fi is awesome for this for several reasons.

1. Having a source inside the house instead of outside is better because you lose, say, 15dB when you go through the wall; this is comparable to what a two-way mirror does to visible light. If you have to illuminate the house from outside using RF energy, you have to deal with much stronger reflections from things outside the house.

2. RF wavelengths that are too short will be badly attenuated by things like walls and doors. You can already notice this with 5GHz 802.11a Wi-Fi; if you have a few walls between you and the AP, the 2.4GHz signal usually works better. The problem gets worse at higher frequencies. (You may have noticed that many walls attenuate visible light, which is RF in the 500THz band, rather strongly.)

3. RF wavelengths that are too long provide much poorer spatial resolution. Outside the near field, your imaging resolution is limited by diffraction to about the wavelength. So you can see a person who's illuminated by the 99.5MHz emissions from your favorite heavy metal station only if their diameter is on the order of 3 m or more, and you can see their movements when they move on the order of 3 m or more. By contrast, 2.4 GHz gives you 120-mm resolution, and 5 GHz gives you 60-mm resolution. For typical humans, these are more useful.

(However, my friend Florian has done good work on passively detecting airplanes using radio illuminations from TV stations, which could be super helpful the next time the US comes to bomb your country, even if he does use Lagrange interpolation instead of B-splines like any normal person would; check it out: https://ieeexplore.ieee.org/document/8115293.)

Also! Having walls be super transparent, as they are at these longer wavelengths, is not entirely an advantage. It makes it harder to distinguish between signals from things in one building and signals from things in another.

If you want to listen to Wi-Fi signal strength changes in real time — including when someone moves around — try https://canonical.org/~kragen/sw/dev3/wifiscan.py. It depends only on Python (3 or recent 2) and PulseAudio. (MacOS hackers, consider upgrading to Linux. Apple's removal of your Esc key shows that they hate you and want you to die.)

I would imagine adding one or more friendly WiFi transmitters as illuminating sources even on the outside would greatly help the accuracy.

Anyone know of a code sample or available project that implements this? Would be a lot of fun to try at home and the office.

I wonder if using Bluetooth is messing with the detection rate. Or maybe tracking the BT emitter makes things even easier.

Modern day Van Eck phreaking!

Does this mean I can finally have a studfinder that works?

didn't everyone see batman?

So, is "magnetic paint" actually a thing, or do we need drywall with Faraday cages?