Scrutable

But, even if it works, isn't there a huge hole in the system - a proportion of the population don't have smart phones and I suspect (without evidence) that many of those are older people - the very ones who are among the most vulnerable ??

lpm wrote: ↑

Tue May 05, 2020 10:30 am

I know it's The Register, which can be sh.t, but this was interesting on the UK NHS branded app. Move to the Isle of Wight to experience it now.

https://www.theregister.co.uk/2020/05/0 ... virus_app/

I came here to post the same thing.

I was wondering how this was supposed to work. My phone battery is a bit shite so I only turn on Bluetooth when I want to transfer data from my Garmin (once a day or so). Looks like even if I leave it on all the time it still won't work so what's the point?

I also have two phones (personal and work) but often only carry one of them (work phone stays in my work bag on evenings/weekends) so, if it worked, I'd need it installed on both phones. However, overnight they will be in close proximity for a long time so would be a contact. That could lead to some interesting skewing of data depending on how prevalent it is to have multiple phones.

Why is this going ahead? (At least in Isle of Wight)

head of NHS digital says

“Even if the take-up rate is 20%, that gives us important insights into how the virus is spreading. At 40 or 50% it will make a big difference.”

https://www.theguardian.com/politics/20 ... ncock-says

How is 20% take up useful? Presumably that's 20% of smartphone owners, which on IoW is going to be, what, 70% of the pop max (taking away kids <14 and half the >65s)?

And the fewer phones there are, the less they are going to be near each other, and even if they are the tech doesn't allow them to speak to each other most of the time. What are they going to be able to find in that kind of ropey data, if they get any at all?

How is this even a good idea? How did they get on in Australia with their (bluetooth-based) app?

the thing with big data is that you can use multiple data sets to de-anonymise data - I notice the UK has a constant number that identifies the user whereas the google / apple proposal the identifier changes each day. Once you crack who the identifier is, (via pattern matching with other data sets) you have access to all their movements.

badger asked how Australia is getting on with it's app. So far, about 4.4m people have downloaded it, so that's about 20% of the population, and most of them downloaded it in the first few days of release. They've only just released the draft legislation so I assume there will be another uptick in downloads once the legislation is passed.

I've got the Australian app on my android phone and I have obviously no clue about what's happening with the data it's collecting but it leaves a little notification up that tells me it's scanning, regardless of whether my phone is locked. I haven't touched it since I installed it so it at least claims to be scanning while in the background. Whenever I turn off Bluetooth, the notification turns to an exclamation mark and asks me to turn Bluetooth back on.

I use Bluetooth headphones a lot of the day anyway so I haven't noticed any effect on battery life.

There are issues here with iPhones around it running in the background or while locked, which they claim are being solved but I don't know how.

The context is very different here too. Modelling suggests that our testing and tracing regime is picking up close to 100% of all cases, so this is an adjunct to that. I don't think the UK is anywhere close to having they level of control over the virus. The app presumably can't replace proper contact tracing unless you could get every human to install it and run it at all times.

I think of it as like the flutracking surveys that I do every week. It will likely track trends that are important and potentially give some useful data on places where transmission is likely - if it's showing that in the two weeks before I tested positive, I was spending lots of time with X number of people and the contract tracers ask me what I was doing at those times, it might help quantify the risks of particular activities. Maybe.

If 40% take up makes a material difference to contact tracing efforts and perhaps to refining social distancing rules, I'm happy to use my relative boringness to help. If I were a member of a vulnerability minority, I might be more cautious, even though I'd be more vulnerable to covid as well as to malign government actions.

As regards the Bluetooth issue (visibility when not active etc.), I've seen it said that the app uses Bluetooth LE, which is designed for use in background and intermittent apps.

https://www.bbc.com/news/amp/technology-52551273

Gfamily wrote: ↑

Wed May 06, 2020 7:43 am

As regards the Bluetooth issue (visibility when not active etc.), I've seen it said that the app uses Bluetooth LE, which is designed for use in background and intermittent apps.

https://www.bbc.com/news/amp/technology-52551273

I have bluetooth on all the time on my phone to keep my smartwatch connected. It is absolutely not a problem. The app (on Android) runs in the notification bar.

"Furthermore, the score is calculated by taking into account all the risky interactions an app user has had over a period of two weeks".

The user will have symptoms by then, if they are ever going to have symptoms. So what's the point of waiting two weeks?

Also, two weeks is a long time in an exponentially-growing outbreak.

Squeak wrote: ↑

Wed May 06, 2020 1:51 am

The context is very different here too. Modelling suggests that our testing and tracing regime is picking up close to 100% of all cases, so this is an adjunct to that. I don't think the UK is anywhere close to having they level of control over the virus. The app presumably can't replace proper contact tracing unless you could get every human to install it and run it at all times.

Yes, the smartphone apps make sense where there is a realistic ambition to test everyone who might be infected. Its difficult to see what good it would do in the UK if someone was notified they are at risk but couldn't get tested.

shpalman wrote: ↑

Wed May 06, 2020 7:59 am

"Furthermore, the score is calculated by taking into account all the risky interactions an app user has had over a period of two weeks".

The user will have symptoms by then, if they are ever going to have symptoms. So what's the point of waiting two weeks?

Also, two weeks is a long time in an exponentially-growing outbreak.

I interpret it to mean that the app only takes the last two weeks of contacts into account in the score. If you don't have symptoms, any contacts before then are irrelevant. If your score indicates you have had serious contact, the app swould presumably recommend taking extra steps to ensure you don't spread it further just in case you are infected but not yet symptomatic.

jaap wrote: ↑

Wed May 06, 2020 10:24 am

shpalman wrote: ↑

Wed May 06, 2020 7:59 am

"Furthermore, the score is calculated by taking into account all the risky interactions an app user has had over a period of two weeks".

The user will have symptoms by then, if they are ever going to have symptoms. So what's the point of waiting two weeks?

Also, two weeks is a long time in an exponentially-growing outbreak.

I interpret it to mean that the app only takes the last two weeks of contacts into account in the score. If you don't have symptoms, any contacts before then are irrelevant. If your score indicates you have had serious contact, the app swould presumably recommend taking extra steps to ensure you don't spread it further just in case you are infected but not yet symptomatic.

If you had a risky contact more than two weeks ago then you should already have been notified about that, once, and then your score should reset based on what you do about it.

ETA: ok what I think it's implying is that your risk is calculated as a moving average over the past two weeks and if at any point it goes over threshold then I suppose you would get the notification.

Thanks Squeak and others. Will be interesting to see how successful the Bluetooth workarounds are. Waking up, finding alien phones and swapping IDs is not the same as being connected to a smartwatch all day that you've already given permission to.

If they can still get something useful even if there's low take up then great, I just hope it's good data for the price.

I've posted a summary of how the app works (not mine) on the Scrutineers FB page.
It's based on the National Cyber Security Centre assessment paper issued on Sunday
https://www.ncsc.gov.uk/files/NHS-app-s ... 20V0.1.pdf

Woodchopper wrote: ↑

Wed May 06, 2020 9:28 am

Squeak wrote: ↑

Wed May 06, 2020 1:51 am

The context is very different here too. Modelling suggests that our testing and tracing regime is picking up close to 100% of all cases, so this is an adjunct to that. I don't think the UK is anywhere close to having they level of control over the virus. The app presumably can't replace proper contact tracing unless you could get every human to install it and run it at all times.

Yes, the smartphone apps make sense where there is a realistic ambition to test everyone who might be infected. Its difficult to see what good it would do in the UK if someone was notified they are at risk but couldn't get tested.

The more I think about this, the more troubling the app seems to me in a society that has an uncontrolled outbreak. If i got a warning for each of ny commutes in the past fortnight, or even once a week, I'd be pretty anxious to get tested. An app that tells me I need to be tested every day is anxiety-inducing in a society where I can get regular tests. In a society where I can't get those tests, that sounds like a recipe for civil unrest.

A regular phone message that makes the government's failures very, very salient seems politically brave.

On technical matters, our senate committee yesterday got told that the problems with iPhones are not yet solved and my native reading seems to hunt they were just need Apple to rebuild some key aspects of their architecture for all their old models and it will be fine. A piece of cake.

Also, the data aren't yet accessible to health authorities so it's not actually doing anything useful yet.

Psythor says it won't work, but someone says it will work, so maybe it will work, but the NHS might abandon it, but maybe they won't:

https://twitter.com/Psythor/status/1258327399391006720

UK starts to build second contact tracing app
NHS team is told to build alternative on system being developed by Apple and Google

The NHS has already begun building a second smartphone app to trace the spread of the coronavirus, after criticism of the first app it launched this week on the Isle of Wight.

The second NHS app will use technology provided by Google and Apple and is being developed “in parallel”, in case politicians decide to make a switch, according to two people familiar with the situation.

Matthew Gould, head of NHSX, the UK health service’s innovation arm, gave the go-ahead to the new project earlier this week.

The decision to build an alternative to the NHS’s original app, which gathers more data in a central database, came after pressure within the government over the technical and ethical issues of its initial approach. 

One person involved said that talks with Apple and Google had intensified in the past few days, noting a sharp change of tack from last week to more “cordial and constructive” discussions “exploring how we might change course”.

The person said that, as testing had continued, the practicalities of making the first app work had become increasingly apparent. He noted a particular problem over its compatibility with the Apple iPhone, as well as broader worries about the implications for battery life. “These technical details end up being quite important,” he said. 

[...]

One person involved in the development of the app said NHSX was now feeling increasing pressure from parliament and privacy campaigners, despite Mr Levy’s assurances. A technical analysis of the app by Privacy International, published on Thursday, found that a loophole in app software could allow authorities access to detailed location data about users in future.

MPs on parliament’s human rights committee also said on Thursday that they had “significant concerns” about the app and called for new legislation guaranteeing data and human rights protections. Harriet Harman, the committee chair, said promises from ministers about privacy were not enough.

NHSX said: “We’ve been working with Apple and Google throughout the app’s development and it’s quite right and normal to continue to refine the app.”

https://www.ft.com/content/446df516-4ec ... pe=nongift

The source code for the UK app has been released and here's a summary of what people who've looked at it say about it

I get that the app is not designed to store location data so that people feel their privacy is protected and are thus more likely to download it. But it strikes me that without location data - or at least the ability to gather it later once permission is given - it's a bit limited.

Sure, when a user gets Covid, anyone else with the app who's been in contact will get a notification. But what about all the contacts that don't have the app? To trace & track those, which may be a sizable portion if the app doesn't take off, you'll need old-fashioned contact tracing and that requires knowing when and where the person went over the last week or so.

I'd be far more confident in my phone remembering everywhere I've been than me.

(I'm pretty cautious about these things & usually have location data turned completely off, but I'd happily turn it on for contact tracing purposes.)

badger wrote: ↑

Tue May 05, 2020 1:50 pm

Why is this going ahead? (At least in Isle of Wight)

head of NHS digital says

“Even if the take-up rate is 20%, that gives us important insights into how the virus is spreading. At 40 or 50% it will make a big difference.”

https://www.theguardian.com/politics/20 ... ncock-says

How is 20% take up useful? Presumably that's 20% of smartphone owners, which on IoW is going to be, what, 70% of the pop max (taking away kids <14 and half the >65s)?

More like 60%. The IoW population is heavily skewed towards the elderly. And even then probably further <60% than one might expect, because while the elderly are often wealthy retirees, the adults there are often in seasonal employment with the low pay that entails.

badger wrote: ↑

Tue May 05, 2020 1:50 pm

And the fewer phones there are, the less they are going to be near each other, and even if they are the tech doesn't allow them to speak to each other most of the time. What are they going to be able to find in that kind of ropey data, if they get any at all?

Well, it can demonstrate herd immunity nicely. Think of people without the app running at all times as the immune

raven wrote: ↑

Tue May 12, 2020 1:51 pm

I get that the app is not designed to store location data so that people feel their privacy is protected and are thus more likely to download it. But it strikes me that without location data - or at least the ability to gather it later once permission is given - it's a bit limited.

Sure, when a user gets Covid, anyone else with the app who's been in contact will get a notification. But what about all the contacts that don't have the app? To trace & track those, which may be a sizable portion if the app doesn't take off, you'll need old-fashioned contact tracing and that requires knowing when and where the person went over the last week or so.

I'd be far more confident in my phone remembering everywhere I've been than me.

(I'm pretty cautious about these things & usually have location data turned completely off, but I'd happily turn it on for contact tracing purposes.)

I bet my phone company are keeping those records of where I've been anyway. Every interaction of my phone with a mast will be recorded & could be used to triangulate my movements if anyone really wanted to.

tom p wrote: ↑

Tue May 12, 2020 3:44 pm

raven wrote: ↑

Tue May 12, 2020 1:51 pm

I get that the app is not designed to store location data so that people feel their privacy is protected and are thus more likely to download it. But it strikes me that without location data - or at least the ability to gather it later once permission is given - it's a bit limited.

Sure, when a user gets Covid, anyone else with the app who's been in contact will get a notification. But what about all the contacts that don't have the app? To trace & track those, which may be a sizable portion if the app doesn't take off, you'll need old-fashioned contact tracing and that requires knowing when and where the person went over the last week or so.

I'd be far more confident in my phone remembering everywhere I've been than me.

(I'm pretty cautious about these things & usually have location data turned completely off, but I'd happily turn it on for contact tracing purposes.)

I bet my phone company are keeping those records of where I've been anyway. Every interaction of my phone with a mast will be recorded & could be used to triangulate my movements if anyone really wanted to.

Information like that has certainly been used (in an anonymised sense) in Italy to get an idea of how many people were making long trips despite the lockdown.

And sometimes it's used by the forces of order to try to find people who've gone missing.

shpalman wrote: ↑

Tue May 12, 2020 5:07 pm

tom p wrote: ↑

Tue May 12, 2020 3:44 pm

raven wrote: ↑

Tue May 12, 2020 1:51 pm

I get that the app is not designed to store location data so that people feel their privacy is protected and are thus more likely to download it. But it strikes me that without location data - or at least the ability to gather it later once permission is given - it's a bit limited.

Sure, when a user gets Covid, anyone else with the app who's been in contact will get a notification. But what about all the contacts that don't have the app? To trace & track those, which may be a sizable portion if the app doesn't take off, you'll need old-fashioned contact tracing and that requires knowing when and where the person went over the last week or so.

I'd be far more confident in my phone remembering everywhere I've been than me.

(I'm pretty cautious about these things & usually have location data turned completely off, but I'd happily turn it on for contact tracing purposes.)

I bet my phone company are keeping those records of where I've been anyway. Every interaction of my phone with a mast will be recorded & could be used to triangulate my movements if anyone really wanted to.

Information like that has certainly been used (in an anonymised sense) in Italy to get an idea of how many people were making long trips despite the lockdown.

And sometimes it's used by the forces of order to try to find people who've gone missing.

It was used in Sweden to identify the members of a criminal gang who used burner phones.
https://books.google.no/books?id=gCqODw ... &q&f=false

Telenor, a big Norwegian telecom company, have already done some work on tracking movement within Norway to predict how Covid-19 might spread. Anonymised by using groups of 20, they say:
https://www.bbc.co.uk/programmes/p0893psr

Also, Click did a nice overview of contact tracing using phones here: https://www.bbc.co.uk/programmes/p08bbhz7

The UK accidentally left secret plans for its COVID-19 contact-tracing app on an open Google Drive

Not sure why they should be secret in the first place but, still, face-elbow.

shpalman wrote: ↑

Thu May 14, 2020 12:13 pm

The UK accidentally left secret plans for its COVID-19 contact-tracing app on an open Google Drive

Not sure why they should be secret in the first place but, still, face-elbow.

I can quite understand how this happened, given the limitations of Google Drive. They presumably used the "Anyone with the link can see" (or comment, or edit) feature for some of the documents, which is in itself fairly secure since the link appears to consist of 42 alphanumeric digits, corresponding to a 256-bit GUID (you would need to wait until the heat death of many universes to generate all the possibilities, even if every atom in every universe was a computer). It seems that they also had a master document with links to other documents whose links hadn't been shared in that way, so that only the few specific accounts who were authorised to see them could follow those links. All secure so far. The problem came when they included the doc with the "Anyone with the link can see" link in among the others. This is pretty hard to catch, because the links look the same whether or not they are "Anyone can access" (you'd have to log in with a spare, "naive" Gmail account on another browser to check it), so I can see how it happened. Maybe Google could add a feature so that "Anyone can access" GUIDs all start with ZZ or something.

The more general problem for any large organisation is that Google Docs and other free apps are extraordinarily tempting (they work reliably, they have cool features --- especially Docs, which sometimes feels miraculous as multiple people work on a text simultaneously, and you may well already be using them outside work, so you don't have to train anyone). It would be essentially impossible for a government to specify and develop anything as smart as that with security on top. Plus, Google isn't interested in making money from software, so you probably couldn't ask them to make a secure version (and they would probably tell you that trying to do so with Javascript in an inherently insecure browser is a bad idea, so now you have to get them to make a secure version of Chrome too).

So for not-especially-secret documents, destined to become part of (or shape discussion around) social policy rather than plan the location of Trident submarines, and where the worst that can happen if they leak is embarrassment, and many of the people working on them don't have much in the way of formal security clearance or training in how to do IT securely in 2020, I don't think it's a particularly huge problem. Certainly it could have happened to anybody, and the alternative would probably have been to have a bunch of people centralised with each other in a smallish room --- news of that leaking out might have been even more embarrassing.

What nobody has explained is, since the trial is on Isle of Wight, do you push Button A or B if you are infected?