Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Rise of Fully Homomorphic Encryption

queue.acm.org

Post reply on HN
Print view View original

21–30 of 132 posts

Re: The Rise of Fully Homomorphic Encryption

#21

Post by vesinisa »

v4dok wrote: post #14
bawolff wrote: post #11

Earlier quoted context omitted.

> FHE gives you NO guarantee about the code that is running on the encrypted data. I can run a leaky AI model or a SELECT * on encrypted data and still get the output. What I can do (and that's assuming there is open-sourced, auditable code) is to make sure that anyone with hypervisor access on that machine cannot dump my data out during processing. I might be misunderstanding but i think this is misleading. Any code…

Depends on what do you mean by "person running the code" If by person you mean the admin of the machine, then yes. If by person you mean the developer of the FHE-based application, then "maybe" If by person you mean the analyst who would in the end order an AI python model to be executed through the FHE-based software on a machine. Then no, that person will in the end get back human-readable results. Be that a model,…

The way I have understood FHE is that any algorithm that would operate on the data would, by definition, be unable to produce any result that was intelligible to anyone except the person holding the original key. At no point during the execution of an FHE algorithm is the data decrypted. The amazing thing is exactly that the code running on the data does not understand the data it is consuming nor the data that it is producing.

Maybe someone who has actually studied homomorphic encryption can chime in.

Re: The Rise of Fully Homomorphic Encryption

#22

Post by ketzu »

v4dok wrote: post #8

I am still amazed by the amount of people that still mix security and encryption in-use with privacy. And truth is the whole privacy/security industry is doing nothing to change that. Take this for example "Valuable insights through AI (artificial intelligence), big data, and analytics can be extracted from data—even from multiple and different sources—all without exposing the data, secret decryption keys, or, if nee…

> A very powerful concept for remote processing, supply chain security, and overall reducing trust; but completely unrelated to privacy.

It is still related to privacy, but the privacy "attacker" is the execution place, allowing for outsourcing of computations and storage without running into data leaks or violations of data protection laws.

Maybe you use a different definition of privacy?

Re: The Rise of Fully Homomorphic Encryption

#23

Post by v4dok »

vesinisa wrote: post #21
v4dok wrote: post #14

Earlier quoted context omitted.

Depends on what do you mean by "person running the code" If by person you mean the admin of the machine, then yes. If by person you mean the developer of the FHE-based application, then "maybe" If by person you mean the analyst who would in the end order an AI python model to be executed through the FHE-based software on a machine. Then no, that person will in the end get back human-readable results. Be that a model,…

The way I have understood FHE is that any algorithm that would operate on the data would, by definition, be unable to produce any result that was intelligible to anyone except the person holding the original key. At no point during the execution of an FHE algorithm is the data decrypted. The amazing thing is exactly that the code running on the data does not understand the data it is consuming nor the data that it is…

Thats true. But it will still produce some data, and that data will be viewed by someone eventually who owns the key to decrypt it. FHE tells you nothing about what this product should be. It could as well be a full copy of the original data.

For example: I run an ML model using FHE on some data I shouldn't have access to in plaintext. The expected outcome of this workflow is a trained ML model on that data. FHE tells me nothing about the quality of this model. It could as well be an overfit model that spits out all the sensitive data.

Re: The Rise of Fully Homomorphic Encryption

#24

Post by zacchj »

If you want to see a practical use case of Fully Homomorphic Encryption (FHE) with Machine Learning.

> https://www.zama.ai/post/titanic-competition-with-privacy-pr...

Its main ambition is to show that FHE can be used for protecting data when using a Machine Learning model to predict outcomes without degrading its performance.

Disclaimer: I'm working at Zama (cited in the article posted).

Re: The Rise of Fully Homomorphic Encryption

#25

Post by adgjlsfhk1 »

arunkant wrote: post #12

Healthcare is stuck in pre 2000 IT technologies because of privacy concerns. I hope that with FHE, Heathcare providers can move to cloud technologies without fear of losing privacy

fhe doesn't solve this at all. if the problem takes little enough power do be run with the, you can run it without cloud compute or fhe easily.

Re: The Rise of Fully Homomorphic Encryption

#26

Post by limbicsystem »

zacchj wrote: post #24

If you want to see a practical use case of Fully Homomorphic Encryption (FHE) with Machine Learning. > https://www.zama.ai/post/titanic-competition-with-privacy-pr... Its main ambition is to show that FHE can be used for protecting data when using a Machine Learning model to predict outcomes without degrading its performance. Disclaimer: I'm working at Zama (cited in the article posted).

Or this! https://pubmed.ncbi.nlm.nih.gov/30641309/

Re: The Rise of Fully Homomorphic Encryption

#27

Post by vesinisa »

bawolff wrote: post #5
motohagiography wrote: post #2

Non-technical comment to consider the conseqeunces of FHE. This is not to diminish the amazing work that has gone into FHE, and the theoretical use cases for FHE in a few fields I've worked in are significant. The challenge I found in working with people who want the data is that they really do just want the data. Examples include government agencies who used made up in-house encryption schemes to get their data shar…

The whole cloud provider using FHE usecase always seemed a bit utopian to me. As you say, most of the time they dont want to provide user privacy, they want your data. Maybe i could imagine some sort of B2B case where there are strong requirements working out, but i struggle imagining it for consumer use cases. Not to mention, if you are outsourcing data computation, presumably its a lot of computation or you would d…

Yeah, I don't think this will work on commercial scale exactly for the same reasons as blockchain is useless for anything outside the illegal niches where you need to avoid the legal banking system.

You may let people store homomorphic data on your servers and even run your algorithms on that data, but you have no way of handling customer complaints or fine-tuning / debugging your service because you can not understand ANY of the customer data you are storing.

Similarly, blockchain sounds like a good idea until you need to reverse a transaction: https://www.pcmag.com/news/cryptocom-sues-woman-after-accide...

Re: The Rise of Fully Homomorphic Encryption

#28

Post by jxf »

v4dok wrote: post #23
vesinisa wrote: post #21

Earlier quoted context omitted.

The way I have understood FHE is that any algorithm that would operate on the data would, by definition, be unable to produce any result that was intelligible to anyone except the person holding the original key. At no point during the execution of an FHE algorithm is the data decrypted. The amazing thing is exactly that the code running on the data does not understand the data it is consuming nor the data that it is…

Thats true. But it will still produce some data, and that data will be viewed by someone eventually who owns the key to decrypt it. FHE tells you nothing about what this product should be. It could as well be a full copy of the original data. For example: I run an ML model using FHE on some data I shouldn't have access to in plaintext. The expected outcome of this workflow is a trained ML model on that data. FHE tell…

Wouldn't doing the same thing without FHE also result in the same problem?

Re: The Rise of Fully Homomorphic Encryption

#30

Post by simiones »

api wrote: post #13
gumby wrote: post #7

This article skips over the elephant in the room, via a couple of casual references to “performance”. You did some experiments with HE in 2019 and it involves orders of magnitude slowdown — thousands of times slower than regular computation. I don’t see this speeding up either.

To make it fast I think we'll need custom silicon. I wonder how many stealth startups exist working on FHE chips?

This isn't a problem that silicon can fix. FHE requires orders of magnitude more operations to be done, and/or more complex operations, to achieve the same results. GP silicon already runs these operations as fast as possible, it's just that there are too many of them, and they are not even parallelizable.
Post reply on HN