[markasoftware]

a bloom filter look up is by hash, and given the relatively small set of words in english, it would be pretty easy for the server to reverse the hash sent to it. Thus a bloom filter wouldn't be very private.

Additionally, a typical spell checker feature is to provide alternative, correct, spellings, rather than just telling you whether a word is correctly spelled.

I bet there's some cool way to do this with zero-knowledge or homomorphic cryptography though!

[notpushkin]

There’s also a way simpler way: send a hash prefix to server, get a list of matches. Google Safe Browsing does this with URLs, for example.

[chuckadams]

haveibeenpwned.com does this for passwords too. I doubt you could make it work for all the smaller words though, let alone offer corrections.

[shakna]

You should be able to do a K-means type thing. Where your query is an entire group, and you grab the field from the chunk locally.

But you might still be able to use some frequency sampling to predict the words used, unless those chunks are very very carefully constructed.

[Sesse__]

> a bloom filter look up is by hash, and given the relatively small set of words in english, it would be pretty easy for the server to reverse the hash sent to it. Thus a bloom filter wouldn't be very private.

The typical use of a Bloom filter is to have it locally as a prefilter, not to send hashes to the server.

[account42]

> I bet there's some cool way to do this with zero-knowledge or homomorphic cryptography though!

The code for which would almost certainly be larger than a fully local dictionary for any human language.

[bmacho]

> a typical spell checker feature is to provide alternative, correct, spellings, rather than just telling you whether a word is correctly spelled.

I personally don't use that one, for me the red underline is enough.