|
Any and all updates regarding the COVID-19 will need a source provided. Please do your part in helping us to keep this thread maintainable and under control.
It is YOUR responsibility to fully read through the sources that you link, and you MUST provide a brief summary explaining what the source is about. Do not expect other people to do the work for you.
Conspiracy theories and fear mongering will absolutely not be tolerated in this thread. Expect harsh mod actions if you try to incite fear needlessly.
This is not a politics thread! You are allowed to post information regarding politics if it's related to the coronavirus, but do NOT discuss politics in here.
Added a disclaimer on page 662. Many need to post better. |
On June 16 2020 17:31 Vivax wrote:
The app would produce a mind boggling amount of false positives. The likelihood of being somewhat in proximity to someone is way higher than being infected in the process.
Probably going to be as 'temporary' as other privacy-breaching measures of the last two decades if it sees widespread use.
Proximity is measured as everything closer than 2m AFAIK. It is a retrospect tool, meaning if someone does test positive, the app checks which devices were in that proximity in the last week and alerts those device to maybe check their health
On June 16 2020 17:32 Acrofales wrote:Show nested quote +On June 16 2020 17:22 Harris1st wrote:On June 16 2020 16:09 Acrofales wrote:On June 16 2020 15:36 Harris1st wrote:The Covid app is available in Germany for iOS and Android. The Corona-Warn-App helps us to determine whether we have had any contact with an infected person which could result in a risk of catching the virus. This way, we can interrupt chains of infection more quickly. The app is a service by the Federal Government. Download and use of the app are entirely voluntary. It is available as a free download in the App Store and from Google Play. www.bundesregierung.de english version This sounds very very Big Brother-like. What are the terms and conditions for the government's data storage and what watchdog is entrusted with ensuring the government doesn't just hold onto that data "just in case". Norway already made a 180 with regards to its Corona app: https://techcrunch.com/2020/06/15/norway-pulls-its-coronavirus-contacts-tracing-app-after-privacy-watchdogs-warning/Without geolocation data, as the German app claims (https://www.thelocal.de/20200402/privacy-mad-germany-turns-to-app-to-track-virus-spread), I know from personal work, that accuracy is a gigantic problem. Contact tracing is not what Bluetooth or wifi were intended for and the algorithms that try it are very very noisy. There are dedicated devices intended for that, but phones are not equipped with that. In fact, all the successful proximity experiments I know of use some form of sensor platform because currently cellphones just aren't good enough. And even without geolocation, if they *do* get this to work, that is still a massive breach of privacy (the government tracking who you were with for how long). Don't get your tinfoil hat in a bunch  Seriosuly though, from what I've read the app was scrutinized by every user safety and data protection firm in the country with no issues (anymore) It has no access to GPS, just BT Already have it installed. If it helps fighting Covid, count me in. And it's not like people value their data. They give it voluntarily to everyone and their mother (facebook, google, apple) But when the government asks "No! they will chip me! Vaccines are bad! Earth is flat! The cake is a lie!" As I said, unless vast strides have been made in the last year or so (last time I tried to do anything in that area), bluetooth is really unreliable as a proximity sensor. It is better as a location sensor for proximity to a fixed base station, but if they aren't using geolocation they obviously aren't doing that. Oh, and it'll also drain your phone battery like nobody's business. So yes, no doubt the privacy watchdogs are now on board, but is the app still at all functional? If it generates more false positives than true positives, how useful is it really? It gives you an upper bound, but a similar upper bound is fairly easily computed without an app.
I can't comment on the technical side as I have no idea how it works.
Will give an 24h update tomorrow about battery life
|
Zurich15365 Posts
On June 16 2020 17:32 Acrofales wrote:Show nested quote +On June 16 2020 17:22 Harris1st wrote:On June 16 2020 16:09 Acrofales wrote:On June 16 2020 15:36 Harris1st wrote:The Covid app is available in Germany for iOS and Android. The Corona-Warn-App helps us to determine whether we have had any contact with an infected person which could result in a risk of catching the virus. This way, we can interrupt chains of infection more quickly. The app is a service by the Federal Government. Download and use of the app are entirely voluntary. It is available as a free download in the App Store and from Google Play. www.bundesregierung.de english version This sounds very very Big Brother-like. What are the terms and conditions for the government's data storage and what watchdog is entrusted with ensuring the government doesn't just hold onto that data "just in case". Norway already made a 180 with regards to its Corona app: https://techcrunch.com/2020/06/15/norway-pulls-its-coronavirus-contacts-tracing-app-after-privacy-watchdogs-warning/Without geolocation data, as the German app claims (https://www.thelocal.de/20200402/privacy-mad-germany-turns-to-app-to-track-virus-spread), I know from personal work, that accuracy is a gigantic problem. Contact tracing is not what Bluetooth or wifi were intended for and the algorithms that try it are very very noisy. There are dedicated devices intended for that, but phones are not equipped with that. In fact, all the successful proximity experiments I know of use some form of sensor platform because currently cellphones just aren't good enough. And even without geolocation, if they *do* get this to work, that is still a massive breach of privacy (the government tracking who you were with for how long). Don't get your tinfoil hat in a bunch Seriosuly though, from what I've read the app was scrutinized by every user safety and data protection firm in the country with no issues (anymore) It has no access to GPS, just BT Already have it installed. If it helps fighting Covid, count me in. And it's not like people value their data. They give it voluntarily to everyone and their mother (facebook, google, apple) But when the government asks "No! they will chip me! Vaccines are bad! Earth is flat! The cake is a lie!" As I said, unless vast strides have been made in the last year or so (last time I tried to do anything in that area), bluetooth is really unreliable as a proximity sensor. It is better as a location sensor for proximity to a fixed base station, but if they aren't using geolocation they obviously aren't doing that. Oh, and it'll also drain your phone battery like nobody's business. So yes, no doubt the privacy watchdogs are now on board, but is the app still at all functional? If it generates more false positives than true positives, how useful is it really? It gives you an upper bound, but a similar upper bound is fairly easily computed without an app.
It's been online for 2 hours and considering there is hardly any Corona anymore in Germany I doubt we will get any true measure on how well it works for tracing. At this point, it's a tool in our belt for a possible second wave. Or a future, non-corona pandemic.
However, of all the technical solutions it's certainly the one that is not at all big brother like. There is no way to use this for surveillance.
Additionally, it is an outstanding example of a public IT project. So much so that I truly couldn't believe it to be true when it was fully released. You can review every piece of the solution and all sources (frontends, backends, everything) here. It's excellently documented:
https://github.com/corona-warn-app/cwa-documentation
|
On June 16 2020 11:19 imBLIND wrote:Show nested quote +On June 16 2020 08:50 soul55555 wrote:
How much longer until covid-19 is over? Moderna"s vaccine just entered phase 2 trials two weeks ago ( www.modernatx.com), as well as AstraZeneca's ( www.fiercebiotech.com) and SinoVac's ( www.fiercepharma.com). So if we're on the fast track route, and ifat least one of them works, we get to phase 3 in the winter...then phase 3 is going to last a year or so before we get to phase 4. In other words, our daily lives is going to be the same for at least another 1.5 years minimum... There's also a plethora of treatments, antibody developments, and other vaccines out there that are being studied.
Oxford's/AstraZeneca's is a combined phase 2 and 3 and they've also been expanding their trial onto other regions like Brazil. I'm not sure they need a phase 4 before starting deployment. They've also stated that results from this could come up between 2 to 6 months from now. Meanwhile they've done agreements with governments, including the US and Europe to supply hundreds of millions of doses by the end of 2020, so I assume they're ramping up manufacturing already at full throttle. It might be possible that there will be vaccines available for frontline hcw and possibly at-risk populations still in 2020?
Portugal has seen new confirmed cases stabilize around 300 per day since late April (!), but on the other hand, deaths have plummeted from 27 per day in late April, to 14 per day in mid May, to 2-3 per day today. It's both frustrating and encouraging, a number of things seem to be happening all at once:
- The health system has become much better at improving outcomes for infected patients;
- We're becoming better at protecting at-risk populations from infection while those getting infected are younger;
- The government has switched from only testing those with symptoms to go out and test entire industries to try to catch most active cases and proactively contain infections.
Authorities are saying they now expect cases to start coming down as they've done a massive effort in finding and containing localized outbreaks. I wish we'd be more like other countries who seem to have contained the virus almost completely, but I'm encouraged by the fact that we seem to have taken the opportunity to get better at fighting it. Hope we're ready for outbreaks in the fall.
|
On June 16 2020 23:09 warding wrote:Show nested quote +On June 16 2020 11:19 imBLIND wrote:On June 16 2020 08:50 soul55555 wrote:
How much longer until covid-19 is over? Moderna"s vaccine just entered phase 2 trials two weeks ago ( www.modernatx.com), as well as AstraZeneca's ( www.fiercebiotech.com) and SinoVac's ( www.fiercepharma.com). So if we're on the fast track route, and ifat least one of them works, we get to phase 3 in the winter...then phase 3 is going to last a year or so before we get to phase 4. In other words, our daily lives is going to be the same for at least another 1.5 years minimum... There's also a plethora of treatments, antibody developments, and other vaccines out there that are being studied. Oxford's/AstraZeneca's is a combined phase 2 and 3 and they've also been expanding their trial onto other regions like Brazil. I'm not sure they need a phase 4 before starting deployment. They've also stated that results from this could come up between 2 to 6 months from now. Meanwhile they've done agreements with governments, including the US and Europe to supply hundreds of millions of doses by the end of 2020, so I assume they're ramping up manufacturing already at full throttle. It might be possible that there will be vaccines available for frontline hcw and possibly at-risk populations still in 2020? Portugal has seen new confirmed cases stabilize around 300 per day since late April (!), but on the other hand, deaths have plummeted from 27 per day in late April, to 14 per day in mid May, to 2-3 per day today. It's both frustrating and encouraging, a number of things seem to be happening all at once: - The health system has become much better at improving outcomes for infected patients; - We're becoming better at protecting at-risk populations from infection while those getting infected are younger; - The government has switched from only testing those with symptoms to go out and test entire industries to try to catch most active cases and proactively contain infections. Authorities are saying they now expect cases to start coming down as they've done a massive effort in finding and containing localized outbreaks. I wish we'd be more like other countries who seem to have contained the virus almost completely, but I'm encouraged by the fact that we seem to have taken the opportunity to get better at fighting it. Hope we're ready for outbreaks in the fall.
There's a ton of countries in the same boat I think.
Germany, Canada, Portugal, Italy are all sitting in the ~350 cases/day region. All of them hit over 1k cases/day at some point, Germany and Italy exceeded that by a lot.
It's possible to control, but eradication seems like a pipe dream without some way of inducing herd immunity, such as a vaccine. The best we can do is protect the elderly and trace/isolate the infected.
Edit:: France is another notable country - they're at ~150 cases/day which is remarkable. Unfortunately it seems to have a very long tail, which means actually eliminating it is very difficult.
|
https://www.bbc.com/news/health-53061281
Looks like the UK's found an existing drug that significantly improves outcomes, Dexamethasone
It cut the risk of death by a third for patients on ventilators. For those on oxygen, it cut deaths by a fifth.
It's an anti-inflammatory so it seems to reduce the chance of patients' immune systems overreacting.
It also seems to have a significant impact on lung damage in surviving patients. It's fairly cheap too, 5 pounds for a 10 day treatment course which is pennies in healthcare.
Edit: article is misleading on costs - going by the quote:
"The treatment is up to 10 days of dexamethasone and it costs about £5 per patient.
"So essentially it costs £35 to save a life.
Pretty promising as a tool in the arsenal of healthcare workers.
|
On June 17 2020 01:59 Lmui wrote:https://www.bbc.com/news/health-53061281Looks like the UK's found an existing drug that significantly improves outcomes, Dexamethasone Show nested quote +It cut the risk of death by a third for patients on ventilators. For those on oxygen, it cut deaths by a fifth. It's an anti-inflammatory so it seems to reduce the chance of patients' immune systems overreacting. It also seems to have a significant impact on lung damage in surviving patients. It's fairly cheap too, 5 pounds for a 10 day treatment course which is pennies in healthcare. Edit: article is misleading on costs - going by the quote: Show nested quote +"The treatment is up to 10 days of dexamethasone and it costs about £5 per patient.
"So essentially it costs £35 to save a life. Pretty promising as a tool in the arsenal of healthcare workers.
This looks significantly more promising than HCQ. Here's to hoping it gets support in the US
|
|
|
On June 16 2020 17:50 zatic wrote:Show nested quote +On June 16 2020 17:32 Acrofales wrote:On June 16 2020 17:22 Harris1st wrote:On June 16 2020 16:09 Acrofales wrote:On June 16 2020 15:36 Harris1st wrote:The Covid app is available in Germany for iOS and Android. The Corona-Warn-App helps us to determine whether we have had any contact with an infected person which could result in a risk of catching the virus. This way, we can interrupt chains of infection more quickly. The app is a service by the Federal Government. Download and use of the app are entirely voluntary. It is available as a free download in the App Store and from Google Play. www.bundesregierung.de english version This sounds very very Big Brother-like. What are the terms and conditions for the government's data storage and what watchdog is entrusted with ensuring the government doesn't just hold onto that data "just in case". Norway already made a 180 with regards to its Corona app: https://techcrunch.com/2020/06/15/norway-pulls-its-coronavirus-contacts-tracing-app-after-privacy-watchdogs-warning/Without geolocation data, as the German app claims (https://www.thelocal.de/20200402/privacy-mad-germany-turns-to-app-to-track-virus-spread), I know from personal work, that accuracy is a gigantic problem. Contact tracing is not what Bluetooth or wifi were intended for and the algorithms that try it are very very noisy. There are dedicated devices intended for that, but phones are not equipped with that. In fact, all the successful proximity experiments I know of use some form of sensor platform because currently cellphones just aren't good enough. And even without geolocation, if they *do* get this to work, that is still a massive breach of privacy (the government tracking who you were with for how long). Don't get your tinfoil hat in a bunch Seriosuly though, from what I've read the app was scrutinized by every user safety and data protection firm in the country with no issues (anymore) It has no access to GPS, just BT Already have it installed. If it helps fighting Covid, count me in. And it's not like people value their data. They give it voluntarily to everyone and their mother (facebook, google, apple) But when the government asks "No! they will chip me! Vaccines are bad! Earth is flat! The cake is a lie!" As I said, unless vast strides have been made in the last year or so (last time I tried to do anything in that area), bluetooth is really unreliable as a proximity sensor. It is better as a location sensor for proximity to a fixed base station, but if they aren't using geolocation they obviously aren't doing that. Oh, and it'll also drain your phone battery like nobody's business. So yes, no doubt the privacy watchdogs are now on board, but is the app still at all functional? If it generates more false positives than true positives, how useful is it really? It gives you an upper bound, but a similar upper bound is fairly easily computed without an app. It's been online for 2 hours and considering there is hardly any Corona anymore in Germany I doubt we will get any true measure on how well it works for tracing. At this point, it's a tool in our belt for a possible second wave. Or a future, non-corona pandemic. However, of all the technical solutions it's certainly the one that is not at all big brother like. There is no way to use this for surveillance. Additionally, it is an outstanding example of a public IT project. So much so that I truly couldn't believe it to be true when it was fully released. You can review every piece of the solution and all sources (frontends, backends, everything) here. It's excellently documented: https://github.com/corona-warn-app/cwa-documentation
That is very cool. They have gone to great lengths to decentralize and ensure anonymity. That does make it more vulnerable to malicious users, but I'll assume the chance of people wanting to vandalize the system are far smaller than people wanting to exploit the data (for instance to stalk someone). So yeah, well done on that front. That leaves accuracy, and I still have serious doubts. This is what they say:
+ Show Spoiler +Procedure
Several times per day, all active Corona-Warn-Apps download the diagnosis keys released on the Corona-Warn-App server and pass them on to the operating system in batches through an interface. The app checks whether any of these received, recorded rolling proximity identifiers match any of the diagnosis keys. If there is a match, this indicates that the user’s smartphone encountered the smartphone of a person who has uploaded a diagnosis key on the day to which the diagnosis key belongs.
In the next step, the app analyzes all the matching rolling proximity identifiers for each diagnosis key, to estimate how long the exposure lasted in total on the day in question and how close the smartphones were to each other on average during the exposure. The distance is calculated from the measured reduction in strength of the Bluetooth signal, which is specified in dBm (decibel-milliwatts). All exposures for a diagnosis key that lasted less than 10 minutes in total (regardless of how close the smartphones came during that time) or during which the smartphones were more than 8 meters (73 dBm) apart on average (regardless of how long the exposure lasted) are discarded as harmless.
NB: In the following, the total of all exposures that belong to a diagnosis key, that is, all exposures over a day between the same two smartphones, is referred to as the “exposure set”.
For the remaining exposures that have not been discarded as harmless, a total risk score is calculated for each exposure set, by multiplying the transmission risk score described above by the days since last exposure value, which is calculated as the time between the day of the last exposure and the current day.
All exposure sets that exceed a certain threshold (the minimum risk score) are considered to be risk exposures. The other exposure sets are discarded as harmless, like the sets that were previously discarded for being too short and/or too distant.
At the same time, the remaining risk exposures are added together to determine how much time exposure took place within a very close range below 1.5 meters (55 dBm) and how much time exposure took place in a close range between 1.5 and 3 meters (63 dBm).
The total calculated time is then cross- calculated against the maximum risk score, the exposure with the highest risk: the time remains unchanged if this risk is estimated as average (for risk exposures), it is extended to one and a half times if the risk is above average, and it is reduced significantly (to around one-sixth) if the risk is below average. As a result, an exposure time of 10 minutes can be extended to more than 15 minutes and an exposure time of 45 minutes can be reduced to less than 10 minutes.
Consequences and Constraints
In the end, a CWA user is notified of an increased risk whenever the risk exposure time calculated as described above amounts to 15 minutes or longer. This notification takes place in the CWA and, at the same time, provides recommendations as to how the user should proceed.
When assessing the times and distances calculated by the CWA, it is important to consider that it is not possible to measure these two parameters precisely. The individually measured times can deviate from the actual exposure time by 5 minutes plus or minus and the calculated distances are approximate values under ideal conditions and the calculated distances are approximate values under ideal conditions, that is, without any impediments between the two smartphones. Even minor impediments, such as a person between the two smartphones or a signal-impeding smartphone case, can cause the distance to appear to be twice as large as it actually is.
Due to privacy considerations, the properties described above can currently only be queried for the total set of all risk exposures at the interface to the operating system, but not for individual risk exposures or exposure by day. As long as the number of new infections remains relatively low, this should not make much of a difference, because it is likely that only very few CWA users will have been exposed to multiple persons who have tested positive within the time frame until they are notified
They describe it well, but the last few paragraphs basically say: "this data is noisy as hell". Depending on the thresholds they configure, it will either give a high number of false negatives (very bad) or a high number of false positives (probably a lot better, but depending on the number may still lead to the platform being useless: an overestimated upper bound can be obtained in other ways too).
What I don't know is how they configured the thresholds. ML? Using what data? Simulation? How? Guessing? The app stands or falls with the calibration of these parameters, and the reason we ended up dropping this as a useful tool is because the Bluetooth signal strength is just *really* unreliable. In addition to the factors they mention, each cellphone has a different quality Bluetooth antenna which may or may not work according to the correct specifications. Motion adds weird artifacts and noise. Atmospheric conditions change the signal strength. As does whether it's in a pocket or a bag or out in the open. Etc. etc. etc. At the end of the day, bluetooth localization is pretty cool, but only works accurately in controlled conditions. And real life Germany is not a controlled environment.
Now I did look into this last a bit of time back, and it's an active field of research. Everybody doing IoT stuff wants this type of thing to work better, and ML algorithms for this type of signal processing are very new and improving fast. Perhaps accuracy has improved to the point of usefulness. But I haven't read of any real breakthroughs either. Which makes me doubt whether this app works better than no app for contact tracing. I didn't see test data either, nor any outline for live field trials.
|
On June 16 2020 17:22 Harris1st wrote:Show nested quote +On June 16 2020 16:09 Acrofales wrote:On June 16 2020 15:36 Harris1st wrote:The Covid app is available in Germany for iOS and Android. The Corona-Warn-App helps us to determine whether we have had any contact with an infected person which could result in a risk of catching the virus. This way, we can interrupt chains of infection more quickly. The app is a service by the Federal Government. Download and use of the app are entirely voluntary. It is available as a free download in the App Store and from Google Play. www.bundesregierung.de english version This sounds very very Big Brother-like. What are the terms and conditions for the government's data storage and what watchdog is entrusted with ensuring the government doesn't just hold onto that data "just in case". Norway already made a 180 with regards to its Corona app: https://techcrunch.com/2020/06/15/norway-pulls-its-coronavirus-contacts-tracing-app-after-privacy-watchdogs-warning/Without geolocation data, as the German app claims (https://www.thelocal.de/20200402/privacy-mad-germany-turns-to-app-to-track-virus-spread), I know from personal work, that accuracy is a gigantic problem. Contact tracing is not what Bluetooth or wifi were intended for and the algorithms that try it are very very noisy. There are dedicated devices intended for that, but phones are not equipped with that. In fact, all the successful proximity experiments I know of use some form of sensor platform because currently cellphones just aren't good enough. And even without geolocation, if they *do* get this to work, that is still a massive breach of privacy (the government tracking who you were with for how long). Don't get your tinfoil hat in a bunch Seriosuly though, from what I've read the app was scrutinized by every user safety and data protection firm in the country with no issues (anymore) It has no access to GPS, just BT Already have it installed. If it helps fighting Covid, count me in. And it's not like people value their data. They give it voluntarily to everyone and their mother (facebook, google, apple) But when the government asks "No! they will chip me! Vaccines are bad! Earth is flat! The cake is a lie!"
The main reason people have disabled the tracking app in Australia is due to its battery usage.With it's constant use of Bluetooth it uses up to 1% battery life per hour.Whether they've figured a workaround for overseas apps for the battery problem I don't know but it is an issue, especially if your battery is already old and declining.
|
1% per hour sounds quite reasonable
|
On June 17 2020 10:37 Acrofales wrote:Show nested quote +On June 16 2020 17:50 zatic wrote:On June 16 2020 17:32 Acrofales wrote:On June 16 2020 17:22 Harris1st wrote:On June 16 2020 16:09 Acrofales wrote:On June 16 2020 15:36 Harris1st wrote:The Covid app is available in Germany for iOS and Android. The Corona-Warn-App helps us to determine whether we have had any contact with an infected person which could result in a risk of catching the virus. This way, we can interrupt chains of infection more quickly. The app is a service by the Federal Government. Download and use of the app are entirely voluntary. It is available as a free download in the App Store and from Google Play. www.bundesregierung.de english version This sounds very very Big Brother-like. What are the terms and conditions for the government's data storage and what watchdog is entrusted with ensuring the government doesn't just hold onto that data "just in case". Norway already made a 180 with regards to its Corona app: https://techcrunch.com/2020/06/15/norway-pulls-its-coronavirus-contacts-tracing-app-after-privacy-watchdogs-warning/Without geolocation data, as the German app claims (https://www.thelocal.de/20200402/privacy-mad-germany-turns-to-app-to-track-virus-spread), I know from personal work, that accuracy is a gigantic problem. Contact tracing is not what Bluetooth or wifi were intended for and the algorithms that try it are very very noisy. There are dedicated devices intended for that, but phones are not equipped with that. In fact, all the successful proximity experiments I know of use some form of sensor platform because currently cellphones just aren't good enough. And even without geolocation, if they *do* get this to work, that is still a massive breach of privacy (the government tracking who you were with for how long). Don't get your tinfoil hat in a bunch Seriosuly though, from what I've read the app was scrutinized by every user safety and data protection firm in the country with no issues (anymore) It has no access to GPS, just BT Already have it installed. If it helps fighting Covid, count me in. And it's not like people value their data. They give it voluntarily to everyone and their mother (facebook, google, apple) But when the government asks "No! they will chip me! Vaccines are bad! Earth is flat! The cake is a lie!" As I said, unless vast strides have been made in the last year or so (last time I tried to do anything in that area), bluetooth is really unreliable as a proximity sensor. It is better as a location sensor for proximity to a fixed base station, but if they aren't using geolocation they obviously aren't doing that. Oh, and it'll also drain your phone battery like nobody's business. So yes, no doubt the privacy watchdogs are now on board, but is the app still at all functional? If it generates more false positives than true positives, how useful is it really? It gives you an upper bound, but a similar upper bound is fairly easily computed without an app. It's been online for 2 hours and considering there is hardly any Corona anymore in Germany I doubt we will get any true measure on how well it works for tracing. At this point, it's a tool in our belt for a possible second wave. Or a future, non-corona pandemic. However, of all the technical solutions it's certainly the one that is not at all big brother like. There is no way to use this for surveillance. Additionally, it is an outstanding example of a public IT project. So much so that I truly couldn't believe it to be true when it was fully released. You can review every piece of the solution and all sources (frontends, backends, everything) here. It's excellently documented: https://github.com/corona-warn-app/cwa-documentation That is very cool. They have gone to great lengths to decentralize and ensure anonymity. That does make it more vulnerable to malicious users, but I'll assume the chance of people wanting to vandalize the system are far smaller than people wanting to exploit the data (for instance to stalk someone). So yeah, well done on that front. That leaves accuracy, and I still have serious doubts. This is what they say: + Show Spoiler +Procedure
Several times per day, all active Corona-Warn-Apps download the diagnosis keys released on the Corona-Warn-App server and pass them on to the operating system in batches through an interface. The app checks whether any of these received, recorded rolling proximity identifiers match any of the diagnosis keys. If there is a match, this indicates that the user’s smartphone encountered the smartphone of a person who has uploaded a diagnosis key on the day to which the diagnosis key belongs.
In the next step, the app analyzes all the matching rolling proximity identifiers for each diagnosis key, to estimate how long the exposure lasted in total on the day in question and how close the smartphones were to each other on average during the exposure. The distance is calculated from the measured reduction in strength of the Bluetooth signal, which is specified in dBm (decibel-milliwatts). All exposures for a diagnosis key that lasted less than 10 minutes in total (regardless of how close the smartphones came during that time) or during which the smartphones were more than 8 meters (73 dBm) apart on average (regardless of how long the exposure lasted) are discarded as harmless.
NB: In the following, the total of all exposures that belong to a diagnosis key, that is, all exposures over a day between the same two smartphones, is referred to as the “exposure set”.
For the remaining exposures that have not been discarded as harmless, a total risk score is calculated for each exposure set, by multiplying the transmission risk score described above by the days since last exposure value, which is calculated as the time between the day of the last exposure and the current day.
All exposure sets that exceed a certain threshold (the minimum risk score) are considered to be risk exposures. The other exposure sets are discarded as harmless, like the sets that were previously discarded for being too short and/or too distant.
At the same time, the remaining risk exposures are added together to determine how much time exposure took place within a very close range below 1.5 meters (55 dBm) and how much time exposure took place in a close range between 1.5 and 3 meters (63 dBm).
The total calculated time is then cross- calculated against the maximum risk score, the exposure with the highest risk: the time remains unchanged if this risk is estimated as average (for risk exposures), it is extended to one and a half times if the risk is above average, and it is reduced significantly (to around one-sixth) if the risk is below average. As a result, an exposure time of 10 minutes can be extended to more than 15 minutes and an exposure time of 45 minutes can be reduced to less than 10 minutes.
Consequences and Constraints
In the end, a CWA user is notified of an increased risk whenever the risk exposure time calculated as described above amounts to 15 minutes or longer. This notification takes place in the CWA and, at the same time, provides recommendations as to how the user should proceed.
When assessing the times and distances calculated by the CWA, it is important to consider that it is not possible to measure these two parameters precisely. The individually measured times can deviate from the actual exposure time by 5 minutes plus or minus and the calculated distances are approximate values under ideal conditions and the calculated distances are approximate values under ideal conditions, that is, without any impediments between the two smartphones. Even minor impediments, such as a person between the two smartphones or a signal-impeding smartphone case, can cause the distance to appear to be twice as large as it actually is.
Due to privacy considerations, the properties described above can currently only be queried for the total set of all risk exposures at the interface to the operating system, but not for individual risk exposures or exposure by day. As long as the number of new infections remains relatively low, this should not make much of a difference, because it is likely that only very few CWA users will have been exposed to multiple persons who have tested positive within the time frame until they are notified
They describe it well, but the last few paragraphs basically say: "this data is noisy as hell". Depending on the thresholds they configure, it will either give a high number of false negatives (very bad) or a high number of false positives (probably a lot better, but depending on the number may still lead to the platform being useless: an overestimated upper bound can be obtained in other ways too). What I don't know is how they configured the thresholds. ML? Using what data? Simulation? How? Guessing? The app stands or falls with the calibration of these parameters, and the reason we ended up dropping this as a useful tool is because the Bluetooth signal strength is just *really* unreliable. In addition to the factors they mention, each cellphone has a different quality Bluetooth antenna which may or may not work according to the correct specifications. Motion adds weird artifacts and noise. Atmospheric conditions change the signal strength. As does whether it's in a pocket or a bag or out in the open. Etc. etc. etc. At the end of the day, bluetooth localization is pretty cool, but only works accurately in controlled conditions. And real life Germany is not a controlled environment. Now I did look into this last a bit of time back, and it's an active field of research. Everybody doing IoT stuff wants this type of thing to work better, and ML algorithms for this type of signal processing are very new and improving fast. Perhaps accuracy has improved to the point of usefulness. But I haven't read of any real breakthroughs either. Which makes me doubt whether this app works better than no app for contact tracing. I didn't see test data either, nor any outline for live field trials.
As someone who spent two years testing bluetooth connection of various phones inside closed space (a car) i have to agree with You. Granted my knowledge is few years old and some progress surely have been made but as far as i know various phones have different implementation of BT stack, they have different drivers, diffrent HW. My educted guess is that this app will work fine on device(s) on which they tested it and be tottally out of sync on others.
|
Yeah 1% per hour is not bad at all.
|
Zurich15365 Posts
I have run the German CWA app for 24 hours now on a 3 year old OnePlus5 and I notice virtually no difference. If the phones battery meter is to be believed the usage is under 1% over 24h. The phone still lasts a good 36 hours with mixed network / wifi usage.
|
Zurich15365 Posts
On June 17 2020 10:37 Acrofales wrote:Show nested quote +On June 16 2020 17:50 zatic wrote:On June 16 2020 17:32 Acrofales wrote:On June 16 2020 17:22 Harris1st wrote:On June 16 2020 16:09 Acrofales wrote:On June 16 2020 15:36 Harris1st wrote:The Covid app is available in Germany for iOS and Android. The Corona-Warn-App helps us to determine whether we have had any contact with an infected person which could result in a risk of catching the virus. This way, we can interrupt chains of infection more quickly. The app is a service by the Federal Government. Download and use of the app are entirely voluntary. It is available as a free download in the App Store and from Google Play. www.bundesregierung.de english version This sounds very very Big Brother-like. What are the terms and conditions for the government's data storage and what watchdog is entrusted with ensuring the government doesn't just hold onto that data "just in case". Norway already made a 180 with regards to its Corona app: https://techcrunch.com/2020/06/15/norway-pulls-its-coronavirus-contacts-tracing-app-after-privacy-watchdogs-warning/Without geolocation data, as the German app claims (https://www.thelocal.de/20200402/privacy-mad-germany-turns-to-app-to-track-virus-spread), I know from personal work, that accuracy is a gigantic problem. Contact tracing is not what Bluetooth or wifi were intended for and the algorithms that try it are very very noisy. There are dedicated devices intended for that, but phones are not equipped with that. In fact, all the successful proximity experiments I know of use some form of sensor platform because currently cellphones just aren't good enough. And even without geolocation, if they *do* get this to work, that is still a massive breach of privacy (the government tracking who you were with for how long). Don't get your tinfoil hat in a bunch Seriosuly though, from what I've read the app was scrutinized by every user safety and data protection firm in the country with no issues (anymore) It has no access to GPS, just BT Already have it installed. If it helps fighting Covid, count me in. And it's not like people value their data. They give it voluntarily to everyone and their mother (facebook, google, apple) But when the government asks "No! they will chip me! Vaccines are bad! Earth is flat! The cake is a lie!" As I said, unless vast strides have been made in the last year or so (last time I tried to do anything in that area), bluetooth is really unreliable as a proximity sensor. It is better as a location sensor for proximity to a fixed base station, but if they aren't using geolocation they obviously aren't doing that. Oh, and it'll also drain your phone battery like nobody's business. So yes, no doubt the privacy watchdogs are now on board, but is the app still at all functional? If it generates more false positives than true positives, how useful is it really? It gives you an upper bound, but a similar upper bound is fairly easily computed without an app. It's been online for 2 hours and considering there is hardly any Corona anymore in Germany I doubt we will get any true measure on how well it works for tracing. At this point, it's a tool in our belt for a possible second wave. Or a future, non-corona pandemic. However, of all the technical solutions it's certainly the one that is not at all big brother like. There is no way to use this for surveillance. Additionally, it is an outstanding example of a public IT project. So much so that I truly couldn't believe it to be true when it was fully released. You can review every piece of the solution and all sources (frontends, backends, everything) here. It's excellently documented: https://github.com/corona-warn-app/cwa-documentation That is very cool. They have gone to great lengths to decentralize and ensure anonymity. That does make it more vulnerable to malicious users, but I'll assume the chance of people wanting to vandalize the system are far smaller than people wanting to exploit the data (for instance to stalk someone). So yeah, well done on that front. That leaves accuracy, and I still have serious doubts. This is what they say: + Show Spoiler +Procedure
Several times per day, all active Corona-Warn-Apps download the diagnosis keys released on the Corona-Warn-App server and pass them on to the operating system in batches through an interface. The app checks whether any of these received, recorded rolling proximity identifiers match any of the diagnosis keys. If there is a match, this indicates that the user’s smartphone encountered the smartphone of a person who has uploaded a diagnosis key on the day to which the diagnosis key belongs.
In the next step, the app analyzes all the matching rolling proximity identifiers for each diagnosis key, to estimate how long the exposure lasted in total on the day in question and how close the smartphones were to each other on average during the exposure. The distance is calculated from the measured reduction in strength of the Bluetooth signal, which is specified in dBm (decibel-milliwatts). All exposures for a diagnosis key that lasted less than 10 minutes in total (regardless of how close the smartphones came during that time) or during which the smartphones were more than 8 meters (73 dBm) apart on average (regardless of how long the exposure lasted) are discarded as harmless.
NB: In the following, the total of all exposures that belong to a diagnosis key, that is, all exposures over a day between the same two smartphones, is referred to as the “exposure set”.
For the remaining exposures that have not been discarded as harmless, a total risk score is calculated for each exposure set, by multiplying the transmission risk score described above by the days since last exposure value, which is calculated as the time between the day of the last exposure and the current day.
All exposure sets that exceed a certain threshold (the minimum risk score) are considered to be risk exposures. The other exposure sets are discarded as harmless, like the sets that were previously discarded for being too short and/or too distant.
At the same time, the remaining risk exposures are added together to determine how much time exposure took place within a very close range below 1.5 meters (55 dBm) and how much time exposure took place in a close range between 1.5 and 3 meters (63 dBm).
The total calculated time is then cross- calculated against the maximum risk score, the exposure with the highest risk: the time remains unchanged if this risk is estimated as average (for risk exposures), it is extended to one and a half times if the risk is above average, and it is reduced significantly (to around one-sixth) if the risk is below average. As a result, an exposure time of 10 minutes can be extended to more than 15 minutes and an exposure time of 45 minutes can be reduced to less than 10 minutes.
Consequences and Constraints
In the end, a CWA user is notified of an increased risk whenever the risk exposure time calculated as described above amounts to 15 minutes or longer. This notification takes place in the CWA and, at the same time, provides recommendations as to how the user should proceed.
When assessing the times and distances calculated by the CWA, it is important to consider that it is not possible to measure these two parameters precisely. The individually measured times can deviate from the actual exposure time by 5 minutes plus or minus and the calculated distances are approximate values under ideal conditions and the calculated distances are approximate values under ideal conditions, that is, without any impediments between the two smartphones. Even minor impediments, such as a person between the two smartphones or a signal-impeding smartphone case, can cause the distance to appear to be twice as large as it actually is.
Due to privacy considerations, the properties described above can currently only be queried for the total set of all risk exposures at the interface to the operating system, but not for individual risk exposures or exposure by day. As long as the number of new infections remains relatively low, this should not make much of a difference, because it is likely that only very few CWA users will have been exposed to multiple persons who have tested positive within the time frame until they are notified
They describe it well, but the last few paragraphs basically say: "this data is noisy as hell". Depending on the thresholds they configure, it will either give a high number of false negatives (very bad) or a high number of false positives (probably a lot better, but depending on the number may still lead to the platform being useless: an overestimated upper bound can be obtained in other ways too). What I don't know is how they configured the thresholds. ML? Using what data? Simulation? How? Guessing? The app stands or falls with the calibration of these parameters, and the reason we ended up dropping this as a useful tool is because the Bluetooth signal strength is just *really* unreliable. In addition to the factors they mention, each cellphone has a different quality Bluetooth antenna which may or may not work according to the correct specifications. Motion adds weird artifacts and noise. Atmospheric conditions change the signal strength. As does whether it's in a pocket or a bag or out in the open. Etc. etc. etc. At the end of the day, bluetooth localization is pretty cool, but only works accurately in controlled conditions. And real life Germany is not a controlled environment. Now I did look into this last a bit of time back, and it's an active field of research. Everybody doing IoT stuff wants this type of thing to work better, and ML algorithms for this type of signal processing are very new and improving fast. Perhaps accuracy has improved to the point of usefulness. But I haven't read of any real breakthroughs either. Which makes me doubt whether this app works better than no app for contact tracing. I didn't see test data either, nor any outline for live field trials.
The solution is protected against vandalism as in people falsely reporting themselves infected if that is what you mean.
Agreed that the beacon data is unreliable and that the system will probably have a high error rate. It will require large scale deployment in a live scenario to see whether it's good enough to contribute to the fight of a pandemic. In that sense it is "unfortunate" that Corona is pretty much beaten in Germany already so there isn't much data out there to test this on. One should also keep in mind that this is not meant to be the end all of all measures to combat Coronavirus or other pandemics, but one tool out of many.
"Fun" fact: The app was supposed to be launched together with relaxed shutdown measures in May. It was delayed precisely because the Government first favored a centralized system, and switched to the now implemented decentralized architecture after it became clear this wouldn't go in Germany. Still, I am still blown away that this project was possible in Germany at all.
|
Can anyone explain how Spain managed to bring down the 7-day average of cases down to something like 300 and deaths to 0? I mean it's quite unlike the other big western european countries like the UK, France, Germany and Italy. Are they just not reporting deaths anymore?
|
On June 17 2020 23:46 warding wrote:
Can anyone explain how Spain managed to bring down the 7-day average of cases down to something like 300 and deaths to 0? I mean it's quite unlike the other big western european countries like the UK, France, Germany and Italy. Are they just not reporting deaths anymore?
Change of reporting methodology, perhaps? I seem to recall various nations having changed their reporting regimes over time.
That said, reporting methodology not being universal kinda makes the global statistics kinda silly to me. How accurate can it be when people report based on different criteria?
|
On June 17 2020 23:46 warding wrote:
Can anyone explain how Spain managed to bring down the 7-day average of cases down to something like 300 and deaths to 0? I mean it's quite unlike the other big western european countries like the UK, France, Germany and Italy. Are they just not reporting deaths anymore?
Looks like they're just not reporting them.
Germany's at ~300 cases a day and low double digit deaths so presumably spain is around there at best. Spain hasn't reported deaths in a week, so likely just empty.
Pretty happy to see NZ has pretty much eradicated it though, and Canada's well on the downwards trajectory. Quebec and Ontario fucked up royally but managed to pull it together. The number of active cases in Canada has dipped below 30k now from a peak of 35k so the first wave has passed peaked. We just have to be careful to ensure a second wave is smaller and keep the border with the states closed until the USA as a whole has the pandemic controlled to the same level as Canada which is likely to not happen.
|
TLADT24920 Posts
^ I think we are likely going to get more cases because I can't see them keeping the border closed for too long lest some of the restaurant that rely on tourists go bankrupt in droves.
|
Importing of cases shouldn't matter that much if you have in place policies to keep Rt around or below 1. And you will have to have those measures in place unless you're an island like NZ and can eliminate the virus. So imported cases will just not result in new outbreaks, they'll barely spread.
|
On June 17 2020 23:46 warding wrote:
Looks like they're just not reporting them.
Germany's at ~300 cases a day and low double digit deaths so presumably spain is around there at best. Spain hasn't reported deaths in a week, so likely just empty.
Germany will "jump" up again!
There is a big slaughterhouse(employees) where they are testing basically everything and for "now" they got over 600 positive hits on about 1000 tests!
https://www.spiegel.de/wirtschaft/covid-19-a-ac3e39fa-3531-492b-90cd-a36cfd1b566c
On the plus side, it seems most of them didn't even know they had it before the testing. So another mark to "this virus is highly infectious but not that dangerous overall!"
|
|
|
|
|
|
EPT
