Bluetooth for contact tracing COVID-19

general topics
21

BT beacons are the same type of RF frontend. They really don’t work that well for distance measurements either. The question they can answer is, is it close?

And it works more the other way around. The beacon beacons. That is, it spits out a regular signal. Your phone knows if it is close to the beacon, not the other way around. The ones I have seen in use are in places like museums. You walk up to an exhibit, phone in hand, and the app you installed at the entrance knows what you are close to, and it displays some information on the object, or maybe plays some audio in your headphones.

A collection of dedicated devices with fixed locations (not beacons, beacons send, they don’t transmit) could triangulate better, sure. But we shouldn’t call it triangulate. It is not the easy triangulation math if you want some accuracy. More like, with the fixed installed devices, you pass many times with a phone via known locations to create a learning set. Then it is black magic time! (Machine learning)

2 Likes
22

The Health Institute here in Norway is rolling out an app soon, possibly even this afternoon. There have been objections from privacy advocates, but without effect. (Note: “two feet” is really “two meters”.)


1 Like
23

Thanks for your ongoing effort to explain BT distance measurement and clearing up my misconceptions about beacons!.

Also:

Thank you for this, too. I have the feeling that this is going to be really interesting, and also going to be interesting breaking it down to people like me, other mutants, and especially decision makers.

Virology and epidemiology already proved hard to communicate. Adding technology and physics… I would pray if I had any believe in that kind of stuff.
Right now, I am really worried.

ETA:

So some regression and some feedback. No problem there. :grimacing:

3 Likes
24

another interesting thing is that the traditional way to do contract tracing involves interviews – often phone calls to people to discuss where they’ve been and when.

only, you need to hire a lot of people to do the interviews, and i guess nobody is looking for work right now… /s

to me… so far… phone based contact tracing seems a bit like electronic voting. sexy and exciting because we are supposed to be living in the future, while the reality is the old school pen-and-paper methods are probably still best.

the us government though – and maybe the general public – would rather give a few private companies big bundles of cash for developing something tech-centric and fragile vs. giving a bunch of individuals each a small amount of cash for doing something routine and reliable.

somewhat related: the supreme court is going to be doing arguments by phone.

whomever thought to use the phone instead of something like zoom deserves plenty of praise imo. reliable and well tested, with no special hardware or training needed.

8 Likes
25

I don’t have much time doing serious research, but a quick look around at scientific papers has had some hits for other techniques besides RSSI.

HOWEVER, everything was either proof of concept or even basic research. Phase shift seemed to be one of the more promising avenues. But to me, as someone who just has some very poor basic understanding of the matter, this is not really clear.

Anyone here willing to dig at ISI, Google Scholar and SciHub?

26

Using the existing BT/BLE hardware in phones or with some newer generation stuff? 5G radios use beamforming techniques so they’re not as constrained by the inverse square law effect. (there is also a direction finding feature in the newer implementation of BLE, but I don’t think it’s widely deployed). One might imagine these coming into play somehow - though I’d find it hard to believe that it would make contact tracing more rather than less effective (increasing the number of uncontrollable variables).

Just from a first principles perspective RSSI is going to fail because radio signal is not a reliable proxy for actual distance / exposure potential, (for that matter even “actual distance” isn’t when considering that walls and multi-level structures exist). What about being in traffic - is every adjacent vehicle a potential exposure source, does isWindowsOpen get evaluated?

What about a scenario where the radio is blocked by granite counter-top, yet the face of the infected person is a couple dozen inches away. The radio that stands for you and the other could have a greater mean-free-path than other radios that are many tens of feet away.

I could see this being useful from a zoomed out population survey perspective (and maybe that’s really all that is actually intended, despite some misdirecting press) - but the utility as a useful marker for individual exposure potential seems extremely low.

1 Like
27

I seriously don’t understand enough to even understand the principles. I looked at some recent papers and found, e.g., phase shift approaches. But the papers were written on dedicated hardware under lab conditions, very much basic research. And I can’t even understand the details, so: no idea what else could be used, but I have the impression we are years, not months, away from a usable technology.

28

An update on the Norway app, which has been deployed for several days:

Apparently, on top of everything else, the part of the system which is supposed to text you if you’ve been near a registered positive is not very secure. I don’t know if that is just bad design in the Norwegian app or is a structural problem with the idea.

1 Like
29

Oh, shit.
Very different kind of problem, but not the last one, we have to assume.

30

@Ratel, I think I disagree. It’s late already, will revisit tomorrow if I can.

Short reply:

at least in my cultural surrounding, plenty of people will take the app not as an indicator, a warning, but as the truth. Hence, no warning, no problems. Which is terrible for false negatives. Warning, problem. Which is very inconvenient for false positives. And false positives will be massive, if you want to reduce false negatives. It is a trade-off. And it will cause massive problems like this.

I think we will be worse off like this.

Additionally, it undermines the credibility of everyone advocating BT contact tracing now. Which is quite bad, since some virologist do, quite possibly trusting the technological advice they got from people who should have the same info @Aciantis posted, but apparently came to a different conclusion.

I still haven’t found one sound explanation why it should word, besides the arguments brought forth in this topic… Which worries me no end.

1 Like
31

One string on their bow that they haven’t mentioned is ultrasonic data transfer. Android uses it with Chromecast pairing to insure that a guest phone trying to give it commands is actually in the same room as the TV.

It’s not perfect either, but the slower speed of sound might help.

1 Like
32

It also definitely does not carry between two cars, or through a wall.

However, this is not what has been proposed so far. The API and all stuff related to contact tracing via smartphones are pitched as using BLE. And I am worried that this cannot work, and people get high hopes. It is so weird to have the feeling to have understood why something is just not what ‘everyone’ is expecting, and see no wider discussion of it. Nothing.

2 Likes
33

Said every climate activist up until about 2017, when tRump’s election (seemingly paradoxically, but not really) caused the climate crisis to break into broad awareness. I know the feeling.

1 Like
34

@LutherBlisset (and others who are interested)
You can easily make some basic experiments if you have access to two mobile phones. Use an app such as “nRF connect” that does basic BLE stuff, and that will show you all BLE devices that are scannable in the area. The app also lets you set up an advertiser. (Basically a BLE beacon. Can be seen and scanned by other people in the area)

Set up one of the phones to be the advertiser. Set it to display its “Complete Local Name” (which you could change). This will help you find it on the second phone, which will use the scanner. The scanner will let you define a filter, where you can input the name of the advertiser so that there is only one device in the list.

Screenshot_20200421-124859
Screenshot_20200421-1248591440×2560 154 KB

The advertiser. Click the EDIT button to make it display “Complete Local Name”. If you wish to change the complete local name, click the pen in the top menu and enter a new name. Otherwise it will just be your phone model.
Screenshot_20200421-124912
Screenshot_20200421-1249121440×2560 159 KB

The edit screen. Click the “ADD RECORD” and select “Complete Local Name”

Then slide the slider on the first screen to turn on your beacon.

Then use the second phone. Open the same app, and select the SCANNER screen at the top. Click SCAN. Stuff will start appearing (how much depends on how many BLE and Bluetooth devices you have in your home). You should find the Advertiser that you set up on the first phone in the list. If you have a long list, you can use the filter function to filter by name.

Screenshot_20200421-131316
Screenshot_20200421-1313161080×2340 200 KB

Look at that! I see my phone. It has a -71 dBm RSSI. (Unfiltered list)

Now you have a setup to experiment with. One thing that is quite noticeable, is that a human body between the phone changes a lot. With one phone on the table and one in my hand I see something like RSSI (40 cm distance) I have around -60 dBm. When I put one phone behind my back (between me and the backrest of the chair) and hold the other in my hand in front of me (about the same distance) the RSSI drops to between -80 dBm and -85 dBm. If we saw that kind of a difference in free-space transmission we would conclude that in the second setup the phone was 10 to 17 times as far away. So for example, a phone in a back pocket will seem farther away in a face to face meeting. And if we walk side by side there is a big difference between a phones in the near pockets, or phones in the far pockets.

(d2/d1 = 10^((Pr1-Pr2)/20) where d2 and d1 are the distances, and Pr1 and Pr2 are the RSSI values in dBm for the two measurements.

(You may have to twiddle the Advertisement type in the setup. Only one of the phones have Bluetooth 5, so I had to set advertisement type to legacy if I wanted that phone to be the beacon. If you struggle, set the older phone to be the beacon, and use the newer for scanning.)

Go forth and play! (In a socially distancing way, of course!)

4 Likes
35

Neat. Even sources available.

Thanks. I believe you without playing around, but will probably do anyway!y

36
2 Likes
37

The thing is: that’s not the problem I wanted to get info on.
The only technical takeaway here is

A Bluetooth signal might be unaware that an infected individual who appeared to be in close proximity to an employee was on the other side of a wall, for example. Nor would the technology be able to register that a person identified as being at risk was wearing protective gear when they had a brush with contagion while shopping

And yes, that is part of the problem described above.

Everyone is discussing privacy, nearly noone if this shit even can work! We are placing our hopes in technology, once again, as usual. Silicon Valley (and Shenzhen, and other areas…) WILL be rubbing their hands in glee. As will Peter Thiel and his PALANTIR cronies, for fucks sake.

Given that we will likely have a terrible amount of false negatives and even higher false positives, this seems not worth handing them our privacy on a silver platter. So can we please get some info how the fuck they think this is going to work despite the technical problems described above? Hello? Hello?

2 Likes
38

I completely agree that this seems like a failure on any practically useful level. Unfortunately don’t have any more authoritative evidence to back it up wrt the tech involved. (I have found some more sources basically reiterating the concerns about false pos/neg rates and the utter indiscerniblability of many highly relevant factors to actual transmission). I fear - as others have suggested here, that the folks proposing this expect to simply hoover up all the data they can and start applying some algorithms and hope something useful falls out. That seems ridiculously naive. I don’t work with this tech in particular, but I do work with arrays of analog sensors that attempt to make sense of unruly multi-dimensional real-world data, and it’s rather tricky to do with accuracy and sensitivity even with highly controlled structures, (which the proposed system wouldn’t have in the least bit).

I think the real nightmare scenario with this BT tracking is that it is easily transformed in a dual-use sense into a super stealthy dissident associate mapper. Once the tech is developed and ubiquitous it becomes trivial to re-purpose even if technically illegal/unconstitutional (that’s what parallel construction is for).

1 Like
39

Australian contact-tracing app leaks telling info and increases chances of third-party tracking, say security folks


(Is this the right thread for this?)

1 Like
40

Well, at least here in Germany, the responsible Ministry of Health made a U-turn and went from centralised data storage to a distributed model after public criticism. Better than nothing.

Well, no. But then, IDK. Maybe yes. We don’t get much new input from anyone who has any idea how this is supposed to work, IRL, with real BLE. So we can as well attract people over and ask everyone to go upthread and read this:

(and this, for people who like to try it out: Bluetooth for contact tracing COVID-19 - #34 by Aciantis)

2 Likes