Teursu144 considering apps, even without internet access, can still communicate with each other (and by this access the internet). I personaly would not take the risk, and just use some other keyboard.
But this is just speculation bascially.
Its probably fine to use like this if your threat model is not that high
dhhdjbd it should be noted that GBoard is most certainly designed to talk to Play Services. If Play Services is installed in the same user profile, GBoard data will likely be sent through that, if it has network access.
Does anyone have any evidence that Gboard does not respect its privacy toggles for disabling data collection?
Its probably fine to use like this if your threat model is not that high
Personally, I don't consider my threat model to be "high". But still I would consider it highly invasive to my privacy if an app collected all my typing data and sent it off to the internet. If that data was published anywhere for other people to see, I imagine I would feel considerably nervous.
I have yet to see any evidence that Gboard does not respect its privacy toggles. Happy to be proven wrong though.
fid02 perhaps it respects that particular toggle which is useless when it works in conjunction with Play services which aid IPC and cookie, identifier and other data collection. Check out work of Doug Leith.
DeletedUser227 Can't see that that source says anything about data collection when Gboard's privacy settings are set to not send data.
Speculating about data extraction through IPC is nothing more than speculation.
This thread basically does discuss this already
While it is technicall possible for gboard/ apps in general to share data via IPC with whoever, i do not think there is any prove that gboard does this.
fid02 Personally, I don't consider my threat model to be "high". But still I would consider it highly invasive to my privacy if an app collected all my typing data and sent it off to the internet. If that data was published anywhere for other people to see, I imagine I would feel considerably nervous.
I ment more like if anything i feel like its some meta data or general analytics.
Furthermore i read that it is not really possible to track ipc communication. And the source code is basically needed for this. Does someone know if this is true?
fid02 the fact remains that IPC exists and metadata is also data, sometimes contextually more important than actual data which is by default protected by app sandbox and permission control.
When weighing options within security and privacy scopes, I would always choose the better option, unless their objectives clash. Then I would always choose privacy first. There are always alternatives. I already use GrapheneOS hardening to maximum reasonable effect.
Last but not least, remember that Google ecosystem was not developed with your best interests at heart but theirs and only theirs so for that purpose you don't always need to run to their defense, they don't need it and they tirelessly work on new methods of implementing tracking technologies that would, surprise, work to their advantage.
This thread basically does discuss this already
And again I think some people seem to reason that:
Thing X is technically possible.
Company Y is generally known for collecting usage data on their users.
Therefore, company Y is doing thing X to collect sensitive data from app Z without consent.
Last but not least, remember that Google ecosystem was not developed with your best interests at heart but theirs and only theirs so for that purpose you don't always need to run to their defense, they don't need it and they tirelessly work on new methods of implementing tracking technologies that would, surprise, work to their advantage.
I'm not surprised that my question about evidence of data collection from a Google app is taken as me somehow "defending" Google. My general impression is that posters in online privacy communities are rarely interested in engaging in nuanced discussion, or to reflect on whether their beliefs are grounded in evidence. This community is usually, sometimes, an exception.
Clearly lots of users here are concerned about Google acting like some kind of malware distributor by distributing an app harvesting data even when the option to opt in is disabled. There are many past posts in this forum expressing this concern. But has anyone actually tried to MITM Gboard's connections to check what data it's sending when the data collection options are disabled? Is there any concrete evidence on this subject at all? Or even concrete indications?
I found a really in depth research paper where they run tests on the data collected by the keyboard. It seems it doesn't log and transmit your keystrokes, but there is extensive metadata being collected.
Anyways, I ended up going with HeliBoard on fdroid. It has most of the features of gboard with better customization and no data collection.
Here is the PDF for the research paper:
https://www.scss.tcd.ie/Doug.Leith/pubs/gboard_kamil.pdf#:~:text=does%20not,ID%2C%20corresponding%20to%20Gboard%E2%80%99s%20and
Teursu144 Here is the PDF for the research paper:
https://www.scss.tcd.ie/Doug.Leith/pubs/gboard_kamil.pdf#:~:text=does%20not,ID%2C%20corresponding%20to%20Gboard%E2%80%99s%20and
That paper concludes that opting out of usage sharing – which is what I think is interesting here – reduces Gboard's data collection to an arguably insignificant amount:
Opting out of this
setting reduced the LATIN_IME logs only to contain a single log entry, which consisted of a
timestamp, the keyboard event "IMS_INPUT_VIEW_STARTED", the Gboard version and
keyboard layout, as well as a couple of other fields which were deemed insignificant.
(p. 31)
Arguably, if someone is concerned that an app that they use shares sensitive data, then it makes sense to not use that app. People like Gboard because it's such a well-made keyboard app, but if you are wary of using it, there are alternatives out there.
