search

bytecodealliance / javy

JS to WebAssembly toolchain
Apache License 2.0
2.06k stars 100 forks source link

Add Crypto HmacSha256 support #696

Open ewalk153 opened 1 week ago

ewalk153 commented 1 week ago

Description of the change

Conforming to the WinterCG crypto Hmac convention, introduce HmacSha256 support. This can be extended to add other crypto functions in the future.

Why am I making this change?

There is a strong need for for function input authentication. HmacSha256 strikes a balance between instruction count and cryptographic safety as a sane starting point.

Checklist

saulecabrera commented 1 week ago

Apologies for the bikeshed here, the spec is a bit hard to navigate, but after investigating a bit more, I believe that what we want is exposing crypto.suble.digest rather than createHmac which is not standard yet (that's why it was not in the doc).

Thankfully I think that this doesn't represent a huge change to the current state of your PR, in fact, I think it might make the implementation a bit easier. The only wrinkle to think about is that crypto.subtle.digest returns a promise, which I believe we can workaround by wrapping the result via: https://docs.rs/rquickjs/latest/rquickjs/struct.Promise.html#method.from_value, without having to actually defer the computation.

saulecabrera commented 1 week ago

The other nice thing is that we can now refer to https://github.com/web-platform-tests/wpt/blob/master/WebCryptoAPI/digest/digest.https.any.js to compliance tests.

ewalk153 commented 1 week ago

Next step: change API signature to:

crypto.subtle.sign({name: "HMAC", hash: "sha-256"}, "my key", "my plaintext")

This aligns with the wintercg spec.

Per convo with @saulecabrera and @jeffcharles

saulecabrera commented 4 days ago

@ewalk153 I've pushed https://github.com/bytecodealliance/javy/pull/696/commits/fa3c661b24c546a6f6af69554eab29da5c8a8a69, see the commit description for details. Additionally, while working on these improvements, I found a bug with our event loop handling code (fixed in https://github.com/bytecodealliance/javy/pull/697), so I'd suggest rebasing once that one lands. I didn't fix all the unit tests in my last commit here, so there's a chance that we'd need to revisit those.