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

Dissection of COVIDSafe (Android): Australian government's contact tracing app

docs.google.com

Post reply on HN
Print view View original

151–160 of 271 posts

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#151

Post by davros »

godelski wrote: post #99
mcannon wrote: post #35

I believe the government, PM and various ministers have said the code will be released. My sources also say exactly the same. They’re obviously operating with extreme urgency to get the app out. For you. Give them a few weeks to clean up code and release it (which is very normal) - but in the meantime, here are some tips: - Turn the HN angry mob mode off - it’s not helpful. We’re all in this together. - Commend the g…

> - Turn the HN angry mob mode off - it’s not helpful. We’re all in this together. I actually believe this is helpful. Just in any democratic setting you want different types of players. You want the "burn the system down" people, because they provide harsh critiques and don't hold back. You want the "okay, but I have reservations" people, because they will push forward but also consider what they are doing (and will…

The Australian governments track record is bad and we should be critical of that. But I have yet to see important problems with this particular app identified. Lets stick with the facts here rather than 'it must be bad'. People have decompiled the kotlin code and so far I haven't seen anyone report anything bad. Am I missing something?

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#152

Post by rstuart4133 »

tastroder wrote: post #2

It's interesting to see these tracing app discussions crop up all over the world at the moment. In Germany it quite literally took dozens of public interest groups, two weeks of media attention, EU guidance and an open letter by hundreds of scientists to make the government switch from central data collection to an acceptable decentralised approach. The amount of misinformation put out by lobby groups in the process…

> is that similar in Australia or is this app primarily driven by the government itself?

I'm an Australian - and from my perspective the answer is "no".

They are doing the best job they can. In this case, the mandarins running the place (we have a West Minister system) look to be very unfamiliar with open sources development practices, and the positive impacts it has software reliability, productivity and the trust you can place in it. But to answer your question - no one selfishly perusing personal agendas or trying to enrich themselves here.

That's not a good excuse for getting it wrong as they have done in this instance of course. But it is just a question of them coming to grips with something they've never bothered to familiarise themselves with. Up until now when they needed a large IT project done, they've just hired IBM at an exorbitant fee. Amazingly, failures brought on by this waterfall style approach of the order of $4B in one instance (and there are many), the collapse of the census and a few weeks ago the collapse of a keystone of their infrastructure never made them consider alternatives. The fact that most successful companies on the planet, the FANG's, main infrastructure is based on open source and it's development model seems to have pass them by unnoticed. So this alternative style of IT development being shoved down their throats is a huge bridge for them to cross.

Here's hoping they make it to the other side :D

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#153

Post by danieltillett »

lbeltrame wrote: post #146
danieltillett wrote: post #137

Earlier quoted context omitted.

Mike my concern as a scientist about this app is it may not help much at this point. If it is only picking up people you spent more than 15 minutes talking to it is going to miss a lot transmission events. Do we have the contact tracing people to actually make use of this data? Even if we did I can’t see how we are going to avoid the need to interview each positive case to find all the people they came into contact w…

> 15 minutes talking to it is going to miss a lot transmission events. Honest question (as a scientist myself): is there any serious non-preprint literature on the time needed for a transmission event (I assume estimates will vary wildly)?

Not that I know of as this is hard to study. What we have is a lot of case studies that transmission can occur in much less time than this including cases where there was no contact between the parties.

The best write up on this I have seen on this topic has been in Quillette [0]. I know Quillette gets attacked here from those on the left, but they do cover a wide range of topics (not all articles I agree with). They are pro-science and generally provide good references.

0. https://quillette.com/2020/04/23/covid-19-superspreader-even...

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#154

Post by shakna »

pca184 wrote: post #95
salan781 wrote: post #80

Earlier quoted context omitted.

Even if the government has the best intentions in this instance, it doesn’t matter. They have already created a set of laws that clearly dictate that this app and this data can be used how ever intelligence communities desire. They have burned all goodwill and trust with the public. It doesn’t matter what they say today unless they repeal AABill etc. Otherwise they’re just saying empty words.

Interpreting legislation without any common law / precedence is difficult. However as a general rule, if there are two laws that are conflicting (such as previous anti-privacy laws vs the proposed safeguards) the most recent enacted law applies, especially if it is specific. So while I’m by no means a fan of the erosion of privacy that this government has done previously, the proposed safeguards would be effective an…

> So while I’m by no means a fan of the erosion of privacy that this government has done previously, the proposed safeguards would be effective and not just empty words (at least legally speaking).

Currently, they are empty words, legally speaking.

The legal text that contains the safeguards is here [0]. It doesn't have most of the safeguards that Hunt announced. They're a pipedream.

For example, the minister said that even in the event of a crime, the data could not be used. However, two parts combine that show actually, they can.

Firstly, possession rather than ownership, controls who can upload data:

> A person must not upload COVID app data from a mobile telecommunications device to the National COVIDSafe Data Store except with the consent of the person who has possession or control of the device.

Secondly, whilst there are controls around who can use that data once it has been uploaded, once it is transferred somewhere for that purpose, there are no restrictions around who can access it once it is outside the data store.

[0] https://www.legislation.gov.au/Details/F2020L00480/Html/Text

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#155

Post by bamboozled »

I'm wondering how the application protects against people running malicious clients works? If the point of this app is to broadcast identifiable information into the public domain, what is stopping others from snooping this information and creating their own tracing DB?

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#156

Post by roenxi »

Daishiman wrote: post #100
discordance wrote: post #45

Earlier quoted context omitted.

I would love to trust them more, but the Australian government does not have a good track record with regards to privacy. Two such recent examples: - Australian's browser history is being provided to law enforcement even though that practice was excluded from the original intent and law [0] - Australia passed laws in 2018 which enable law enforcement to compel tech companies into inserting backdoors into their softwa…

The Australian government is not a single monolothic institution. It's a set of hundreds of thousands of people with different agendas and intentions. You cannot assign singular agency to the entire government; rather we have to go case-by-case to understand the implications of programs like these.

That is a little like saying the human body isn't a single organism because, say, the digestive tract operates with its own agency and intention outside of conscious control. It isn't correct; the politicians in the government have a very high level of agency around what ends up being fed into the institutions and what agendas and intentions are allowed to rise to power.

This isn't the step that gets us to a dystopian future, but it is so cheap and convenient for government to take programs like this and expand them every single time there is a crisis that it may as well be assumed to be coming if people don't kick up a stink each and every time.

We don't need perfect safety. We've can't have perfect safety. Having perfect, technologically enforced safety will create systems that will become corrupted and evil with a high, high likelihood. I don't want the government to have the ability figure out who I'm talking too at all; I'd rather we went in the exact opposite direction of this app and put legal barriers in place to them even asking. COVID-19 is horrible, but it will pass. This tracking strategy will not.

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#157

Post by rstuart4133 »

tgsovlerkhgsel wrote: post #132

Does this randomize the Bluetooth address too? I saw the README (from the dissection) mention a function that hides the name "so the other side only gets the address", which would defeat the entire purpose of rotating identifiers. If it does randomize the Bluetooth address, does it use a separate identifier, and if so, does it rotate both at the same time? Otherwise, you can use an identifier that changes at time 1 t…

It's OpenTrace [0]. OpenTrace is GPLv3, and is based on a published specification that's not too difficult understand. The fact that it is dervived from OpenTrace and they haven't published the source is the whole basis of this story.

To answer your direct questions:

- randomize Bluetooth addresses: I expect not, as that would screw any existing bluetooth connections, like headsets.

- does it use a separate randomised identifier: yes.

On Android at least you would be foolish to trust it without a verifiable chain of trust from the source to the binary you are running. It has two things that matter greatly: your true name, and your precise location. There is nothing physically preventing them from uploading your whereabouts every 10 minutes to a server - so you have to trust the binary doesn't do that. Right now we only have their word [1]. Whether you care enough above the sort of information it could leak to need to trust it is a different question. But if you do care, you would be a fool to do so without a verifiable chain.

A verifiable chain of trust means:

- source starts from a trusted origin. (It does: opentrace)

- there is a cryptographically signed audit trail showing how they change it to get to its current state. (The original is in github, so that's possible).

- they publish the source before deployment. (The two points above means someone inspecting the result only has to look at the changes, not the entire thing).

- they use a reproducible build.

[0] https://github.com/opentrace-community

[1] Right now I'm sure they are good for their word. Move on 24 months and if you still have it installed, then based on their past history I would not trust them as far as I could kick them.

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#158

Post by rstuart4133 »

SyneRyder wrote: post #81

I've been running COVIDSafe on Android for most of today, Samsung's Battery monitor is showing 3% battery use by COVIDSafe after 6 hours. I guess that's about a 10-12% battery hit over a full day, but at least it's using less battery than Spotify or TuneIn or Pocket Casts when they were in use with the screen off. So we're not talking Pokemon Go levels of battery drain here. It works fine in the background on Android…

The battery problem seems to be an Apple thing. iOS can shut the app down if it is in the background, and will do so if the battery is getting low. One solution is to keep it in the foreground - but that does chew power.

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#159

Post by lbeltrame »

danieltillett wrote: post #153
lbeltrame wrote: post #146

Earlier quoted context omitted.

> 15 minutes talking to it is going to miss a lot transmission events. Honest question (as a scientist myself): is there any serious non-preprint literature on the time needed for a transmission event (I assume estimates will vary wildly)?

Not that I know of as this is hard to study. What we have is a lot of case studies that transmission can occur in much less time than this including cases where there was no contact between the parties. The best write up on this I have seen on this topic has been in Quillette [0]. I know Quillette gets attacked here from those on the left, but they do cover a wide range of topics (not all articles I agree with). They…

Thanks. An interesting and fascinating read. I hope more research is done on the topic: from my semi-untrained (I work in oncology, not in virology, although I studied the immune system for about four years in my career) there far more models around than experimental data (which is obviously far harder to gather correctly).

Re: Dissection of COVIDSafe (Android): Australian government's contact tracing app

#160

Post by askvictor »

lbeltrame wrote: post #146
danieltillett wrote: post #137

Earlier quoted context omitted.

Mike my concern as a scientist about this app is it may not help much at this point. If it is only picking up people you spent more than 15 minutes talking to it is going to miss a lot transmission events. Do we have the contact tracing people to actually make use of this data? Even if we did I can’t see how we are going to avoid the need to interview each positive case to find all the people they came into contact w…

> 15 minutes talking to it is going to miss a lot transmission events. Honest question (as a scientist myself): is there any serious non-preprint literature on the time needed for a transmission event (I assume estimates will vary wildly)?

I beleive there is also a lack of literature on how useful a contact tracing app would be in the first place. Even for lack of studies, releasing the modelling and assumptions would be useful. Some considerations:

* Possible increase in false positives bogging down testing regime? * Surface-based (i.e. location-based) contact events (e.g. elevator button) * Effectiveness on health-care workers, who themselves will likely be in contact with infected people a lot, perhaps despite having sufficient PPE.

I get that it won't be perfect, and doesn't need to be perfect, but I'd at the least like to see some modelling to see what they've considered, and how likely useful the app will be.

Post reply on HN