Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

mmWave radar, you won't see it coming

joshhorne.com

Post reply on HN
Print view View original

61–70 of 192 posts

Re: mmWave radar, you won't see it coming

#61

Post by hornej »

parsimo2010 wrote: post #57
transpute wrote: post #49

> Because mmWave radar doesn’t process visual light, what it sees is not personally identifiable. With machine learning improvements, identification and more are possible. From a 2021 paper on IEEE 802.11bf Wi-Fi Sensing, https://arxiv.org/pdf/2103.14918.pdf > Indeed, it has been shown that SENS-based classifiers can infer privacy-critical information such as keyboard typing, gesture recognition and activity tracking…

Right on. The whole "mm wave radar is detail rich but gives privacy" line is a contradiction. Any data that is detailed enough to be useful is detailed enough to be a privacy concern. While it might not be the same as leaking nude photos from someone's camera, radar data is something that needs to be carefully thought out. Radar data from a car can provide information about people's driving behavior even without loca…

My main point is the data is anonymous at source. I think having control over how that data is correlated back to an individual is important

Re: mmWave radar, you won't see it coming

#62

Post by smegsicle »

parsimo2010 wrote: post #57
transpute wrote: post #49

> Because mmWave radar doesn’t process visual light, what it sees is not personally identifiable. With machine learning improvements, identification and more are possible. From a 2021 paper on IEEE 802.11bf Wi-Fi Sensing, https://arxiv.org/pdf/2103.14918.pdf > Indeed, it has been shown that SENS-based classifiers can infer privacy-critical information such as keyboard typing, gesture recognition and activity tracking…

Right on. The whole "mm wave radar is detail rich but gives privacy" line is a contradiction. Any data that is detailed enough to be useful is detailed enough to be a privacy concern. While it might not be the same as leaking nude photos from someone's camera, radar data is something that needs to be carefully thought out. Radar data from a car can provide information about people's driving behavior even without loca…

> leaking nude photos from someone's camera

whatever happened to that meme about common ccd cameras picking up uv images that make out details through eg swimsuits, but hopefully filtered and removed in software before making it to the images?

Re: mmWave radar, you won't see it coming

#63

Post by hornej »

krasin wrote: post #51
samatman wrote: post #46

Earlier quoted context omitted.

I have no opinion on your reply, I was discussing technical feasibility in response to someone claiming it was infeasible. The normative claims aren't germane or particularly interesting to me.

By not exposing anyone I meant reconstructing an image. Identification is a different story and there are multiple ways on how to achieve that (heartbeat classification is an obvious one). If you have ideas how to reconstruct an image with 4 receivers / 2-3 transmitters, please, buy a dev kit from TI (they cost https://www.mouser.com/c/?q=mmwave ) and show us a demo.

I have a lot of experience with TI's IWR6843AOP chip and dev kit. It has 3 transmitters and 4 receivers and it'd be impossible to create a facial reconstruction with it.

Re: mmWave radar, you won't see it coming

#64

Post by dheera »

smegsicle wrote: post #62
parsimo2010 wrote: post #57

Earlier quoted context omitted.

Right on. The whole "mm wave radar is detail rich but gives privacy" line is a contradiction. Any data that is detailed enough to be useful is detailed enough to be a privacy concern. While it might not be the same as leaking nude photos from someone's camera, radar data is something that needs to be carefully thought out. Radar data from a car can provide information about people's driving behavior even without loca…

> leaking nude photos from someone's camera whatever happened to that meme about common ccd cameras picking up uv images that make out details through eg swimsuits, but hopefully filtered and removed in software before making it to the images?

Why not just fix the swimsuits? Seems like a much more straightforward solution.

Re: mmWave radar, you won't see it coming

#65

Post by dathinab »

> mmWave radar doesn’t process visual light, what it sees is not personally identifiable

that is not how "personally identifiable" works

as long as you can (re-)identify a person with it it's personally identifiable data. Even if identification is done by using a mmWave radar and some simple ML.

Re: mmWave radar, you won't see it coming

#66

Post by fooblaster »

I've been somewhat disappointed that the learning material around mimo radar processing that can be done with popular TI automotive radar chips (and others) is somewhat limited after you get through with the absolute basics.

TI's documents are a good starting point: https://training.ti.com/sites/default/files/docs/mmwaveSensi...

But it doesn't really go into the different schemes for sending out orthogonal waveforms from the transmitters, the design of 2d arrays, much of the basic signal processing (windowing, matched filtering), etc.

I was pleased to see https://hforsten.com/mimo-radar-antenna-arrays.html pop up on hacker news quite a while ago, which was very interesting. Goes into how to plot the beam width for a mimo array, with some nice python code. I want more, but I'm afraid I won't get it unless I read through a phd thesis. Any ideas/tips?

Re: mmWave radar, you won't see it coming

#67

Post by yjftsjthsd-h »

krasin wrote: post #5
idontwantthis wrote: post #3

> Because mmWave radar doesn’t process visual light, what it sees is not personally identifiable. This makes it an ideal choice for monitoring where a camera would not be appropriate. Is 5mm resolution not enough to identify a face? What’s the difference between a 3d model of a person constructed from high res radar and a black and white photo? The radar sounds even more identifiable. Edit: Couldn’t you construct a m…

I worked with mmWave radars previously, but not professionally. In my opinion, it would be quite challenging to extract enough features from the current generation of mmWave radars, such as TI IWR1642 ([1]) which have around 4 receivers and 2-3 transmitters, because the incoming data, while having a lot of temporal resolution is very limited in spatial resolution. With greater number of antennas something like what y…

...so what I'm hearing is "hard today so probably trivial in 5 years"? Or is this an exception to the usual march of technological progress?

Re: mmWave radar, you won't see it coming

#68

Post by hornej »

transpute wrote: post #49

> Because mmWave radar doesn’t process visual light, what it sees is not personally identifiable. With machine learning improvements, identification and more are possible. From a 2021 paper on IEEE 802.11bf Wi-Fi Sensing, https://arxiv.org/pdf/2103.14918.pdf > Indeed, it has been shown that SENS-based classifiers can infer privacy-critical information such as keyboard typing, gesture recognition and activity tracking…

I agree that if all of a sudden our routers started mapping out the world like in the Batman movie that would be a bad thing.

Re: mmWave radar, you won't see it coming

#69

Post by dathinab »

transpute wrote: post #49

> Because mmWave radar doesn’t process visual light, what it sees is not personally identifiable. With machine learning improvements, identification and more are possible. From a 2021 paper on IEEE 802.11bf Wi-Fi Sensing, https://arxiv.org/pdf/2103.14918.pdf > Indeed, it has been shown that SENS-based classifiers can infer privacy-critical information such as keyboard typing, gesture recognition and activity tracking…

Hm, I the EU due to GDPR this services are required to make you opt in to them.

Does that mean many of this applications are outright illegal in the EU, due to not having a good way to handle the case where the user doesn't opt-in?

Re: mmWave radar, you won't see it coming

#70

Post by avidphantasm »

dathinab wrote: post #65

> mmWave radar doesn’t process visual light, what it sees is not personally identifiable that is not how "personally identifiable" works as long as you can (re-)identify a person with it it's personally identifiable data. Even if identification is done by using a mmWave radar and some simple ML.

Yes, it seems much too early to say that this is privacy preserving just because visible wavelengths aren’t involved.
Post reply on HN