Live data from Hacker News

Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

github.com

21–30 of 39 posts

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#21
> Smartphone companies like Apple and Google, in an attempt to better their word prediction algorithms, as well as collect potentially profitable information on what their users were typing, began submitting and aggregating all their user's key-taps to their servers, violating user's privacy in the process.

Neither Google's nor Apple's keyboards send your keystrokes to the server, that would have been too outrageous, even for them. Gboard is kind of leaky, as it uses "federated learning" where the keyboard finetunes a prediction model locally, then the results are sent to the server and merged there; this is literally a long-term profile of what (and how) you type, and can probably be used to identify you elsewhere. But even then it doesn't send your keystrokes directly.

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#22
post #4

For iOS i found this excellent T9 keyboard recently: https://apps.apple.com/se/app/type-nine-t9-keyboard/id926008... Also work without any network access, lookups are on-phone

I think you know, but just to be clear: this one isn't T9, it's something IMO much better. I've used the keyboard they based this on and it's easily the best input method available on a phone once you learn it.

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#24

> Smartphone companies like Apple and Google, in an attempt to better their word prediction algorithms, as well as collect potentially profitable information on what their users were typing, began submitting and aggregating all their user's key-taps to their servers, violating user's privacy in the process. Neither Google's nor Apple's keyboards send your keystrokes to the server, that would have been too outrageous,…

The model updates are not persistently stored and are not accessed individually. The system can only access aggregated averages of updates.

Here is a paper that describes the cryptography protecting user data https://eprint.iacr.org/2017/281

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#25
I'm missing some information on how this relates both to GKOS and particularly KeyBee! I did a survey of these different input methods many years ago[1]. Shame things haven't developed much since!

[1]: https://two-wrongs.com/rethinking-text-input-on-touchscreens...

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#26
post #23

It's unfortunate that holding down backspace does not keep deleting characters (or better yet delete word by word like Swiftkey does)

Delete word is "swipe left on backspace". The autorepeat is definitely something I'd like to see too.

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#27
post #2

Neat idea, shame about all this misleading information though. The website very strongly implies (1) the only way to predict words is by uploading every keystroke to central server,(2) every keyboard other than theirs do that and (3) the only way to get privacy is to download their product. This is all complete lies of course. The prediction database is small enough that you can bundle it with app and run all predict…

> The prediction database is small enough that you can bundle it with app and run all prediction locally.

I don't think this is necessarily true. It really depends on what you're trying to predict. That said, I would agree that it should be 100% optional for your keyboard to have network access.

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#28
post #2

Neat idea, shame about all this misleading information though. The website very strongly implies (1) the only way to predict words is by uploading every keystroke to central server,(2) every keyboard other than theirs do that and (3) the only way to get privacy is to download their product. This is all complete lies of course. The prediction database is small enough that you can bundle it with app and run all predict…

It's quite misleading, actually. GBoard uses Federated Learning [0] so that models are trained locally on the device and keywords are not sent to the central server. Both Google and iOS use Federated Learning, Federated Evaluation & Tuning, Differential Privacy, and other privacy-preserving AI techniques and/or on-device ML to avoid sending user data to the server. Both have done so publicly since 2017 and have published many papers and talks on the subject.

0: https://arxiv.org/abs/1602.05629

Re: Thumb-Key: A privacy-conscious Android keyboard, made for your thumbs

#29
post #24

> Smartphone companies like Apple and Google, in an attempt to better their word prediction algorithms, as well as collect potentially profitable information on what their users were typing, began submitting and aggregating all their user's key-taps to their servers, violating user's privacy in the process. Neither Google's nor Apple's keyboards send your keystrokes to the server, that would have been too outrageous,…

The model updates are not persistently stored and are not accessed individually. The system can only access aggregated averages of updates. Here is a paper that describes the cryptography protecting user data https://eprint.iacr.org/2017/281

TIL, thanks! I totally missed this part and just assumed it was just another profiling tool. (it's Google, it has to be, somehow...)
Post reply on HN