Intel Demos Chip To Compute With Encrypted Data

https://spectrum.ieee.org/fhe-intel

93 points · 11 comments · view on lemmy.world

11 Comments

krispyavuz@lemmy.world · 42 pts · 176d (6 replies)

And we are SURE Intel will not hold a backdoor for the supposedly encrypted data

Tehdastehdas@piefed.social · 20 pts · 176d (2 replies)

That's impossible, the keys are not there.

https://en.wikipedia.org/wiki/Homomorphic_encryption

ATS1312@lemmy.dbzer0.com · 1 pts · 175d (1 reply)

How do you get useful "predictive analytics" in healthcare without breaking that encryption? This smells like snake oil from cloud-vendors when I don't have rigorous math proofs cited in the wikipedia page.

FooBarrington@lemmy.world · 3 pts · 174d

Homomorphic Encryption is a well-known field of research in cryptography. Honestly, if you are capable of understanding the proofs, you don't need them to be listed on Wikipedia.

PabloSexcrowbar@piefed.social · 2 pts · 175d (1 reply)

Couldn't you technically figure out what the data is by doing a bunch of operations on it and observing the result? I'm not a math or security guy by any stretch but this seems almost too good to be true.

FooBarrington@lemmy.world · 1 pts · 174d

No, you couldn't.

NoSpotOfGround@lemmy.world · -2 pts · 176d

Yes, it's guaranteed!

Oh, you said "not"...

xylogx@lemmy.world · 20 pts · 175d

Homomorphic encryption is pretty wild, you can sort and search data without decrypting it.

SkybreakerEngineer@lemmy.world · 7 pts · 175d (2 replies)

Cute, but it looks like there aren't actually any FIPS approved homomorphic protocols. That's a big risk for someone like a bank to take on

Archr@lemmy.world · 7 pts · 175d

FIPS updates/approvals always take a long time.

Cocodapuf@lemmy.world · 2 pts · 174d

Trust the math. You don't need to accept risk if you can check the math yourself.

That's what cryptography is about.