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

Machine Learning on Encrypted Data Without Decrypting It

juliacomputing.com

Post reply on HN
Print view View original

81–90 of 122 posts

Re: Machine Learning on Encrypted Data Without Decrypting It

#81

Post by fspeech »

orbifold wrote: post #41
ChrisRackauckas wrote: post #20

This blog post reminds me of the "Machine Learning Systems are Stuck in a Rut" paper [1], where they mentioned: > It is hard to experiment with front end features like named dimensions, because it is painful to match them to back ends that expect calls to monolithic kernels with fixed layout. On the other hand, there is little incentive to build high quality back ends that support other features, because all the fron…

Julia always seemed great on paper and definitely is a strong candidate for replacing Matlab. But whenever I tried using it, the user experience seemed much more broken than python or c++. It just seems way easier to structure and work on a python + c++ project than it is to structure and work on a Julia project. A moderately sized sane c++ code base compiles and runs faster than whatever gymnastics Julia performs to…

The particular attraction to me that Julia advertises is that it is much easier to read other people's Julia code. I recently tried Julia and found this to be quite true. It still lacks some critical features I need in the ML library but I can definitely see its attraction to users who want to understand the libraries they use.

Re: Machine Learning on Encrypted Data Without Decrypting It

#82

Post by winrid »

KenoFischer wrote: post #69
winrid wrote: post #64

I'm assuming this uses similar encryption algorithms that you can use to do queries on encrypted data?

I'm assuming you're referring to private information retrieval ( https://en.wikipedia.org/wiki/Private_information_retrieval ), which is from the same field of research, but may or may not use the same techniques.

Never looked into it. Just heard about it recently. Very cool.

Re: Machine Learning on Encrypted Data Without Decrypting It

#83

Post by ReidZB »

falcolas wrote: post #65
KenoFischer wrote: post #4

Earlier quoted context omitted.

Just to clarify, homomorphic encryption does not expose patterns. At every point in the computation the ciphertexts are computationally indistinguishable from random. The result of evaluating the ML model will be an encrypted prediction that you then need to send back to whoever encrypted the data (or more precisely whoever has the key - doesn't need to be the same person) so they can decrypt and use the prediction.

So, you're training a statistical model - teaching it to recognize certain patterns - on data that is somehow wholly without patterns? Even with a certain amount of noise in the individual data points, if you're given enough data to train a statistical model to identify traits in the ciphertext you also probably have enough data to break the encryption. Also, something can be statistically random and still have patte…

> I believe there's a reason that homomorphic encryption has not been broadly accepted as an allowed standard.

Well, yes... Practical homomorphic encryption is cutting-edge research, and standards bodies like NIST aren't going to deal with an area like this until it's much closer to "solved" (by which I mean much more efficient, with more practical applications, widely used and scrutinized schemes, etc.)

Re: Machine Learning on Encrypted Data Without Decrypting It

#84

Post by personjerry »

This technique is deeply flawed.

You can't do this effectively without outside knowledge they should not have. They are in fact using outside knowledge, specifically that the encrypted data is in the form of images. Without that knowledge, you wouldn't know which ML techniques to use! Additionally, remember that feature engineering is a big part of what makes ML effective at all, and that certainly cannot be done on encrypted data (the feature engineering you need to do depends on what the data looks like).

Re: Machine Learning on Encrypted Data Without Decrypting It

#85

Post by zionic »

rehasu wrote: post #76

Think about what encryption should do. Think about what Machine Learning should do. The only thing you can do with ML on encrypted data is show where encryption needs to be improved. Or maybe there is a way to create ML models that produces output which only someone with the correct (private) key can understand.

Precisely. Any encryption that allows an attacker to significantly reason about its underlying plaintext is broken.

Re: Machine Learning on Encrypted Data Without Decrypting It

#86

Post by heavenlyblue »

falcolas wrote: post #72
KenoFischer wrote: post #68

Earlier quoted context omitted.

> Once the model is trained, I can run this on any data encrypted with key X and find out if the data contains a cat (with some degree of accuracy). I have no way of telling information outside the tags provided on the training data, but it still gives me, a person without the encryption key, the ability to identify any feature that's tagged in the training set on any un-tagged production set. I think you're missing…

Predictions are a tag, and a separate confidence value. A fairly finite set. Cat. Not a Cat. Large quantities of small finite sets are anathema to encryption.

If you map a binary variable into a, let's say, 128-bit number (non-linearly) where a mapping is unique for every encryption key used, there's nothing making it an anathema for encryption.

I am not sure whether this is achievable practically though.

Re: Machine Learning on Encrypted Data Without Decrypting It

#87

Post by sthatipamala »

personjerry wrote: post #84

This technique is deeply flawed. You can't do this effectively without outside knowledge they should not have. They are in fact using outside knowledge, specifically that the encrypted data is in the form of images. Without that knowledge, you wouldn't know which ML techniques to use! Additionally, remember that feature engineering is a big part of what makes ML effective at all, and that certainly cannot be done on…

If I don't even know what data type my input is, what meaningful output can the system give? I wouldn't consider that cheating.

Also, you can do feature engineering on a training set that you collect, as long as that is similar to the distribution of end user inputs. That's a pretty standard ML workflow.

Re: Machine Learning on Encrypted Data Without Decrypting It

#88

Post by littlestymaar »

With Homomorphic Encryption, the data owner encrypt data X, and give C(X) it to the owner of the function f applies it to C(X), which gives f(C(X)), which equals to C(f(X)) and return it to the data owner, which can decipher it to f(X). The owner of the function know neither the initial data nor the result, and the owner of the data doesn't know f.

There is also Functional Encryption, a different technique solving a similar (but different) goal. With F.E., the owner of the data must know the function f, but the party performing the calculation get directly the result f(X).

Re: Machine Learning on Encrypted Data Without Decrypting It

#89

Post by personjerry »

sthatipamala wrote: post #87
personjerry wrote: post #84

This technique is deeply flawed. You can't do this effectively without outside knowledge they should not have. They are in fact using outside knowledge, specifically that the encrypted data is in the form of images. Without that knowledge, you wouldn't know which ML techniques to use! Additionally, remember that feature engineering is a big part of what makes ML effective at all, and that certainly cannot be done on…

If I don't even know what data type my input is, what meaningful output can the system give? I wouldn't consider that cheating. Also, you can do feature engineering on a training set that you collect, as long as that is similar to the distribution of end user inputs. That's a pretty standard ML workflow.

The entire point of encrypted data is you don't know what's in there. What they're claiming is they've worked around this problem, but they haven't, because they're subtly incorporating outside information: specifically that they know what the data looks like already. From a true third-party looking at encrypted data, you would have no idea.

What does your statement about feature engineering have to do with encrypted data? Maybe I can clarify with an example: You can't transform a waveform into more usable features with a fourier transform if the waveform is encrypted.

Re: Machine Learning on Encrypted Data Without Decrypting It

#90

Post by heavenlyblue »

KenoFischer wrote: post #74
falcolas wrote: post #72

Earlier quoted context omitted.

Predictions are a tag, and a separate confidence value. A fairly finite set. Cat. Not a Cat. Large quantities of small finite sets are anathema to encryption.

I'm still only guessing at the objection, but if it helps, the encryptions are randomized, so two different encryptions of the same value do not have the same ciphertext.

Not all encryption algorithms are susceptible to known-cyphertext attacks.
Post reply on HN