Author Topic: Bluetooth Low Energy is unsuitable for COVID-19 contact tracing, say inventors  (Read 9131 times)

0 Members and 1 Guest are viewing this topic.

Offline cdevTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 7350
  • Country: 00
In particular, its problems will lead to many false positives and negatives they say.

https://theintercept.com/2020/05/05/coronavirus-bluetooth-contact-tracing/

"What the large print giveth, the small print taketh away."
 

Offline m98

  • Frequent Contributor
  • **
  • Posts: 634
  • Country: de
Which is kind-of obvious. Getting reliable meter-level ranging from single RSSI measurements with dissimilar devices, antenna orientations, from a body-worn device, in a multipath environment with a multitude of obstacles.
It just sounds like a dead-end idea from the beginning. Maybe its good enough to tell if people where together in the same room/building.
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15794
  • Country: fr
Indeed. It's flawed.

It's not even 100% good enough to tell if people were nearby at some point. There are tons of reasons why in some particular conditions, detection would fail, and especially in crowded places with a lot of radio devices.

I have no idea of the average "detection" % in real-life settings.

In a past project, we had implemented BLE and depending on conditions, it could get finicky. Detecting ALL devices in range could take sometimes several minutes, which may not be compatible with people moving around.
« Last Edit: May 11, 2020, 02:04:22 pm by SiliconWizard »
 

Offline jogri

  • Frequent Contributor
  • **
  • Posts: 405
  • Country: de
Well, if you would combine BLE with GPS and the gyrometer/inertial measurement unit you could get rather accurate histograms of a persons position. If another device gets registered via BLE you wait until the signal drops out and then compare the (more or less) exact position of both persons over the time they were in proximity. Should yield way better results (you basically eliminate the RSSI/walls issue), but you would have to share your movement/exact position with everyone in your proximity-> won't happen as it would be a data security/privacy nightmare.

One possible way to improve the current system would be to install BLE beacons in highly frequented areas that just announce their exact position (+transmission strength)->the app could determine the RSSI of an object that is x meters away if it compares the location of the beacon against its own GPS signal-> triangulation, that would partly counteract the problem that your body is reducing the signal strength in some directions.
 

Offline mark03

  • Frequent Contributor
  • **
  • Posts: 749
  • Country: us
In most western countries, GPS has been ruled out over privacy concerns.  I think the BLE-alone approach is the only alternative seriously being considered.  (Actually deploying it would require a country with at least a basic minimum level of faith in government, and functional high-level decision making, i.e. not the U.S.)

Clearly BLE is not perfect, but it doesn't have to be perfect.  It only has to be better than the [achievable] alternatives.

Anyway, the thread title is an unfair characterization (sorry, characterisation ;)) of the article, the title of which was
Quote
The Inventors of Bluetooth Say There Could Be Problems Using Their Tech for Coronavirus Contact Tracing
 
The following users thanked this post: I wanted a rude username

Offline Syntax Error

  • Frequent Contributor
  • **
  • Posts: 584
  • Country: gb
This comment is from the UK NHSX Covid Tracking app on github.com

/nhsx/COVID-19-app-Android-BETA/blob/master/app/src/main/java/uk/nhs/nhsx/sonar/android/app/DeviceDetection.kt

Code: [Select]
    // TODO: re-consider this condition after we understand how devices advertise
    // without supporting MultiAdvertisement feature (isMultipleAdvertisementSupported == false)
    // This allows for them to run without crashing but probably not advertise as expected.
    // See: https://stackoverflow.com/a/32096285/952041
    // Here is part of the previous device detection logic that blocked Huawei P Smart phone.
    // After commenting it out we are not sure if the phone's advertising works
    // with our Bluetooth scanning process:
    //  || (bluetoothAdapter.isEnabled && !bluetoothAdapter.isMultipleAdvertisementSupported)
    // TODO: We need a real device to test Bluetooth scanning if isMultipleAdvertisementSupported == false
    // TODO: We need analytics to identify number of devices that fall into this bucket
    // and if they scan correctly
 

Offline vodka

  • Frequent Contributor
  • **
  • Posts: 518
  • Country: es

With a court warranty applying to everyone, the government can just triangulate everyone and the past history, down to tens of meters of accuracy, by cell station logs.

That is literally, the state declares a State of exception. The  goverment legally  can't  intervene the  communications massively without  a court order

Also, your payment info like POS devices used with cards and mobile payments, can all be pulled to pretty accurately track everyone's path.

This is IMO much less invasive than having people to install apps, and considering the virus transmits over thin air, there's no sense precisely pin pointing the exact location anyway.

Yes, it  has matter that we  know the routes with accuracy, because it is not same put 10 persons in quarantine than 10000 persons.


 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
You don't need BLE, you don't need GPS, actually you don't need any apps running on phones or whatsoever.
With a court warranty applying to everyone, the government can just triangulate everyone and the past history, down to tens of meters of accuracy, by cell station logs.

... no Warrants shall issue, but upon probable cause, supported by Oath or affirmation, and particularity describing the place to be searched, and the persons or things ...

So what crime did everybody commit for which probable cause exists?

https://en.wikipedia.org/wiki/Warrant_(law)#History

We won that war so we do not have to put up with General Warrants anymore.
 
The following users thanked this post: Gregg, SiliconWizard

Offline cdevTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 7350
  • Country: 00
But maybe 14 year old jennifer Zhang's high school science project makes all this fuss unnecessary.

We certainly owe it to ourselves to find out. (See attached PDF) 

She won a prize for it.

You don't need BLE, you don't need GPS, actually you don't need any apps running on phones or whatsoever.
With a court warranty applying to everyone, the government can just triangulate everyone and the past history, down to tens of meters of accuracy, by cell station logs.

... no Warrants shall issue, but upon probable cause, supported by Oath or affirmation, and particularity describing the place to be searched, and the persons or things ...

So what crime did everybody commit for which probable cause exists?

https://en.wikipedia.org/wiki/Warrant_(law)#History

We won that war so we do not have to put up with General Warrants anymore.
"What the large print giveth, the small print taketh away."
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 7043
  • Country: nl
Impressive science project ... but I don't think you can't get that dose from an ingested medicine. Even with an inhaled one I'm kinda scared what that amount of it would do to everything else but the virus.
 

Offline Mr. Scram

  • Super Contributor
  • ***
  • Posts: 9821
  • Country: 00
  • Display aficionado
... no Warrants shall issue, but upon probable cause, supported by Oath or affirmation, and particularity describing the place to be searched, and the persons or things ...

So what crime did everybody commit for which probable cause exists?

https://en.wikipedia.org/wiki/Warrant_(law)#History

We won that war so we do not have to put up with General Warrants anymore.
Snowden's revelations ruined that dream a bit.
 

Offline Syntax Error

  • Frequent Contributor
  • **
  • Posts: 584
  • Country: gb
Here's a thought for Bluetooth contact tracing. If you can use the RSSI to calculate distance as per the equation*
Distance = 10 ^ (( MeasuredPower – RSSI)/(10 * N)) , it would be possible to penalise people for breaching social distancing rules as their recorded RSSI is too high. Or at least give them a proximity notificaction!

* Source: How to Calculate Distance from the RSSI value of the BLE Beacon
https://iotandelectronics.wordpress.com/2016/10/07/how-to-calculate-distance-from-the-rssi-value-of-the-ble-beacon/

« Last Edit: May 11, 2020, 08:19:18 pm by Syntax Error »
 

Offline cdevTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 7350
  • Country: 00
Could you explain a bit more?

This paper on RESV + MERS said that the dose was attainable and sustainable by multiple dosing.
(see attached picture here is the original paper
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5307780/ )

Resveratrol is proposed elsewhere as likely good for most people who are at high risk for COVID-19. Papers have been written just about that.


Impressive science project ... but I don't think you can't get that dose from an ingested medicine. Even with an inhaled one I'm kinda scared what that amount of it would do to everything else but the virus.

Also, weights of a substance that is inhaled are smaller not larger than can be ingested. Large amounts of resveratrol are toxic but that is a dose that nobody would take. According to this other study the amounts needed would be sustainably taken. With the caveat that I am not a doctor and have no idea if this would work, and although it is entirely reasonable to theorize that smaller doses might have the desired result of reducing the severity of infection so as to make it less dangerous, but perhaps not completely preventing it, its also a notoriously unpredictable situation that nobody has any right to or should try to second guess.

Something similar has been done with virulent pseudorabies virus and piglets.

See https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6164078/

Actually, a number of papers have replicated this kind of experiment in a number of different animals and viruses.

Note: I have NO idea if this would work against COVID-19.

I would assume it won't. Take precautions! I would rather make a fool of myself wearing masks and gloves and washing my hands six times a day than get a potentially fatal disease.

Also MERS is a different SARS-like betacoronavirus.

To my knowledge COVID-19 still has not been tested with resveratrol. I've run out of entities to call about this. If you live in other countries, you might consider asking them if they have tested it. just to jog their memories that they should.

Tell them we may be dropping the ball.

Point them at this:

Sirtuins Are Evolutionarily Conserved Viral Restriction Factors-
https://mbio.asm.org/content/mbio/5/6/e02249-14.full.pdf
« Last Edit: May 11, 2020, 08:45:12 pm by cdev »
"What the large print giveth, the small print taketh away."
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 20355
  • Country: gb
  • 0999
I admit I’ve not read it yet but thought I’d chime in anyway. It doesn’t have to be perfect and no system can be, as someone might pick it up from an object. The app just has to ensure people who might have it get tested, without there being too many false positives to overwhelm the system. That’s all that’s needed to keep the R number below one, without isolating everyone.
 

Offline Buriedcode

  • Super Contributor
  • ***
  • Posts: 1718
  • Country: gb
Seems a bit click baity to me.  The app doesn't claim to be super accurate, and if someone has their diagnosis confirmed it can at least show if they've been self isolating, or been in relative close proximity with lots of people.  And of course, it can't know if someone has been in contact with someone without a phone.

It isn't meant to completely track everyone, the question is - is it better than nothing? 

No-one is claiming an app can magically know everyone who will get it, but it should at least help modify peoples behavior by informing them if they've been near a confirmed case.  They'll probably have to tweak it to reduce the false positives, and there will be plenty of false negatives but some feedback is better than none.
 

Online nali

  • Frequent Contributor
  • **
  • Posts: 732
  • Country: gb
AIUI There are essentially 2 types of systems being proposed - the centralised and the peer-2-peer.

The centralised system has everyone's phone just reporting its position to a server then everything is worked out from there - which obviously raises privacy concerns.

The P2P relies on BLE detection which has several flaws. Apart from the signal itself being subject to environmental influences, different phones report signals in different ways e.g. I had a Samsung J3 which reported the same signal some 40dB higher than all my other phones. Also, not all phones and OS versions support advertising.

I don't know how they've considered scenarios like I could be on a bus or train, and a few feet away from someone at a stop or station. Or cars in adjacent lanes in traffic, building with partition walls etc etc etc..



 

Offline cdevTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 7350
  • Country: 00
Seems a bit click baity to me.  The app doesn't claim to be super accurate, and if someone has their diagnosis confirmed it can at least show if they've been self isolating, or been in relative close proximity with lots of people.  And of course, it can't know if someone has been in contact with someone without a phone.

It isn't meant to completely track everyone, the question is - is it better than nothing? 

No-one is claiming an app can magically know everyone who will get it, but it should at least help modify peoples behavior by informing them if they've been near a confirmed case.  They'll probably have to tweak it to reduce the false positives, and there will be plenty of false negatives but some feedback is better than none.


Bluetooth low energy has enough range to be received much further away than what is a safe distance for COVID-19. Especially if there are metal objects nearby which may act as RF reflectors or blockers.

For example, lets assume that 40 people are riding on a bus and that 35 of them have cell phones with BTLE turned on. The interior of the bus, desite the existence of windows, will be likely to concentrate the RF so that all the phones receive more signals from one another even if they are on opposite ends of the bus. 

So, as far as proximity, even if only four people are near enough to somebody with covid-19 to possibly have picked up the virus, it would be likely that far more than four people might show as having been close, even people far away.  Most of us have been in a large round room, or one with a curved roof. A so called whispering gallery, You can stand on one side of the round room and somebody else can stand at its exact opposite nodal point and hear you whisper. Effects like that are common with RF.

An RTLSDR can be useful in showing how much UHF signals vary as you move an antenna around.
"What the large print giveth, the small print taketh away."
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 20355
  • Country: gb
  • 0999
Seems a bit click baity to me.  The app doesn't claim to be super accurate, and if someone has their diagnosis confirmed it can at least show if they've been self isolating, or been in relative close proximity with lots of people.  And of course, it can't know if someone has been in contact with someone without a phone.

It isn't meant to completely track everyone, the question is - is it better than nothing? 

No-one is claiming an app can magically know everyone who will get it, but it should at least help modify peoples behavior by informing them if they've been near a confirmed case.  They'll probably have to tweak it to reduce the false positives, and there will be plenty of false negatives but some feedback is better than none.


Bluetooth low energy has enough range to be received much further away than what is a safe distance for COVID-19. Especially if there are metal objects nearby which may act as RF reflectors or blockers.

For example, lets assume that 40 people are riding on a bus and that 35 of them have cell phones with BTLE turned on. The interior of the bus, desite the existence of windows, will be likely to concentrate the RF so that all the phones receive more signals from one another even if they are on opposite ends of the bus. 

So, as far as proximity, even if only four people are near enough to somebody with covid-19 to possibly have picked up the virus, it would be likely that far more than four people might show as having been close, even people far away.  Most of us have been in a large round room, or one with a curved roof. A so called whispering gallery, You can stand on one side of the round room and somebody else can stand at its exact opposite nodal point and hear you whisper. Effects like that are common with RF.

An RTLSDR can be useful in showing how much UHF signals vary as you move an antenna around.
I don't see how any of that really matters. Over the last 7 weeks we've had blanket social distancing: assuming everyone is infected and physically isolating everyone from one another, as much as possible to prevent transmission. This was a very blunt tool which cut R below 1, at huge economic cost. What the government should have done right from the beginning, was more testing and contact tracing and testing, as has been the case in Taiwan and South Korea, which haven't had so much social distancing. Now the infection rate has been reduced to manageable levels, the idea is we can sharpen our a approach and work towards only isolating those who are most likely to have the virus. Ideally we should be able to test everyone who's most likely to have the virus, then they can be free to return to work.

The app is just another tool to help with contact tracing. Now we only need to isolate everyone on the bus, rather than the whole country and because it's a relatively small number of people, it's more feasible to test everyone one of them and the friends and family of those who test positive. In the UK, the app will be voluntary and I believe it will still have a positive effect, even if not everyone uses it. If it's proven to be effective enough, then there should be some incentives for people to use it such as being able to jump long queues to enter supermarkets more quickly and perhaps money off certain items such as toilet paper. ;)

Other measures which many may consider to be big-brotherish will be required. For example, if a nightclub is opened, everyone going will need to show photo-ID on the door, then if anyone feels unwell afterwards and test positive, everyone who visited the club on that night, along with anyone they've been in close contact with, will have to be tested and self-isolate if appropriate. A similar thing will need to be done in other situations where social distancing is impossible, such as crowded public transport. The app should only be a small part of the programme of testing and treatment/isolation of the infected.

If governments invest hugely in contact tracing and targeted testing, we can all get back to work and play fairly quickly and the tax revenues will easily pay for it.

To those who are concerned about big brother: they have a right to be, but don't forget many of these measures will be costly and governments won't want to spend money on them when they're no longer needed. I can't even see China bothering to waste money on recording the name and address of everyone entering a crowded bar.
« Last Edit: May 12, 2020, 07:19:36 pm by Zero999 »
 

Offline cdevTopic starter

  • Super Contributor
  • ***
  • !
  • Posts: 7350
  • Country: 00
Accuracy matters a great deal. IMHO.

If they are using a technology that cannot be accurate enough according to its own developers, maybe its not the right technology!
 
« Last Edit: May 12, 2020, 08:40:41 pm by cdev »
"What the large print giveth, the small print taketh away."
 

Offline Buriedcode

  • Super Contributor
  • ***
  • Posts: 1718
  • Country: gb
Accuracy matters a great deal. IMHO.

If they are using a technology that cannot be accurate enough according to its own developers, maybe its not the right technology!

The only alternative would be to use absolute location - GPS - which, as pointed out raises obvious privacy concerns.  Also, I don't believe Apple allow for continuous location tracking on any app.  I'm sure there will be a version that uses location tracking, but users would have to opt in, and I'm sure the powers-that-be are pretty hesitant because of an inevitable backlash.

The phrase "accurate enough" implies there is a known level of accuracy desired.  There isn't.  Even if somehow current smart phones had the ability to measure proximity to other phones with mm accuracy - at what point would it flag up "possible contagion contact" ?  Would it be a sliding scale?  Would that only be inside enclosed spaces?  All the app really needs to do (for now) is to know whether or not the phones owner has been tested, and the result, and how often they have come into reasonably close proximity to other users that aren't in their household.  I'm sure the devs working on this aren't making any kind of claims about efficacy - right now they're busy ironing out bugs and it will be an on-going experiment.
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 7043
  • Country: nl
How about using BLE for a conservative estimate and then using 20-22 kHz audio for ranging when people are near?

Or maybe even just use audio period. Autocorrelation of a chirp signal don't care about multipathing.
« Last Edit: May 12, 2020, 09:12:05 pm by Marco »
 

Offline jogri

  • Frequent Contributor
  • **
  • Posts: 405
  • Country: de
How about using BLE for a conservative estimate and then using 20-22 kHz audio for ranging when people are near?

Or maybe even just use audio period. Autocorrelation of a chirp signal don't care about multipathing.

The "RSSI" would be absolutely terrible as soon as you put your phone into your pocket. Plus, having a microphone that is listening all the time is actually worse than having GPS running 24/7 when it comes to privacy. You need at least 60% of the population to use this app, if you have to admit that you (the government) have basically created a bug that is spying on people when they use the app no one will use it.
« Last Edit: May 12, 2020, 09:23:09 pm by jogri »
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 7043
  • Country: nl
The "RSSI" would be absolutely terrible as soon as you put your phone into your pocket.
Autocorrelation with a chirp doesn't get you RSSI, it gets you multipath insensitive time of flight for the closest path. As for SNR, repetition, repetition, repetition. (Ping-pong ranging makes it a little more complex than that, but it's just audio ... almost nothing in audio is processor intensive any more by today's standards.)
Quote
Plus, having a microphone that is listening all the time is actually worse than having GPS running 24/7 when it comes to privacy.
No it's not.

With an open source app and tons of researchers verifying the integrity of the appstore version you can be certain it doesn't do anything with the data except ranging and identification and that that data does not leave the phone.

With GPS data in a central database however no such verification is possible, you're reliant on trust of the database maintainers.
« Last Edit: May 12, 2020, 09:35:55 pm by Marco »
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 20355
  • Country: gb
  • 0999
Accuracy matters a great deal. IMHO.

If they are using a technology that cannot be accurate enough according to its own developers, maybe its not the right technology!

The only alternative would be to use absolute location - GPS - which, as pointed out raises obvious privacy concerns.  Also, I don't believe Apple allow for continuous location tracking on any app.  I'm sure there will be a version that uses location tracking, but users would have to opt in, and I'm sure the powers-that-be are pretty hesitant because of an inevitable backlash.

The phrase "accurate enough" implies there is a known level of accuracy desired.  There isn't.  Even if somehow current smart phones had the ability to measure proximity to other phones with mm accuracy - at what point would it flag up "possible contagion contact" ?  Would it be a sliding scale?  Would that only be inside enclosed spaces?  All the app really needs to do (for now) is to know whether or not the phones owner has been tested, and the result, and how often they have come into reasonably close proximity to other users that aren't in their household.  I'm sure the devs working on this aren't making any kind of claims about efficacy - right now they're busy ironing out bugs and it will be an on-going experiment.
Yes and I doubt the developers themselves are epidemiologists. Any small degree of accuracy is better than the current situation. Even if it errs on the side of caution, generating some false positives, it will still mean fewer people, than now, will need to be isolated to control the spread.
 

Offline jogri

  • Frequent Contributor
  • **
  • Posts: 405
  • Country: de
When i said "RSSI" i meant loudness of the recieved signal (that's why it is in quotation marks). Fabric dampens the volume quite a bit, and combined with the fact that the microphones were never designed to be used at this frequency range the SNR will be abysmal.

Btw, good luck telling the general pulic that an app that is developed by the government and will most likely be verified by researcher that are working for the government (universities etc) can be trusted... You just need one guy telling everyone that said app is used to spy on people via the microphones. And you have to confirm that "yes, the app is constantly listening via the microphone BUT...". I don't think it would matter what comes after the "but", as soon as you say it the damage is done.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf