Show HN: Building an End-to-End Encrypted Shazam with Homomorphic Encryption

zama.ai

59 points·zacchj··8 comments

This is awesome but with a mere 1000-song database it would be simpler just to run the whole thing on the client. How well could the approach scale? (eg. To a billion song DB?)

Yes for now, it's 1000 song, which is already awesome if you think about it, no? As it's like 300 ms, one can increase the DB size by a few order of magnitude, certainly. It will scale to billions of songs thanks to hardware accelerators, which are coming. One can google and see that there is a bunch of companies (small or large) working on accelerating FHE computations.

> This is awesome but with a mere 1000-song database it would be simpler just to run the whole thing on the client

If you have a Google Pixel phone running the stock OS, you already have this! https://support.google.com/pixelphone/answer/7535326?hl=en#z...

Nice idea, but do I need E2E to identify a song? Seems like a very low threat model for a malicious attacker to know my wife needs that Elton John song.

Yes, but I think it illustrates that FHE has the power to safeguard users' privacy for any app that require mic access... !

I'd be curious to the training time (per training example) of the logistic regression model in FHE.

Everything is open-source, you can have a look yourself, and experiment! https://github.com/iamayushanand/Concrete_Shazam/blob/main/M...

Here, the training is not done on encrypted values: the songs are public, what is secret is which song(s) you like

Really cool!

Reply on news.ycombinator.com