General Questions on Big Data and AI privacy

alexander

1/17/24, 7:53 AM

All,

Recently, I came across a question on privacy for big data and AI.

IMO, big data privacy focuses on "anonymization" aspect where sensitive informatino such as Personal Identitfiable Information should be protected, while AI privacy focuses on "raw data stealth" where raw data should not be inferred or derived from the training and inference processes.

Just want to know other's thought. All comments are welcomed!

1 + 4

data-privacy

Titanlord

1/17/24, 8:59 AM

Simplified there are two approaches for privacy in crypto: The easy one is k-anonymity, and the advanced one is differential privacy. Are you aware of them?

alexander

1/19/24, 1:05 AM

@Titanlord, I know these privacy preserving approaches. But this post is not going to learn solutionts. At first, I want to differentiate these two problems. More precisely, does big data privacy belong to AI privacy or vice versa or what's the exact relationship between these two?

kodlu

1/20/24, 3:33 PM

since both "big data" and AI are fuzzy and fungible concepts I don't think a precise answer could be pinned down. Maybe someone will prove me wrong.

DannyNiu

2/19/24, 12:20 PM

I’m voting to close this question because the relationship between cryptography and the topic of this question is (as I've felt) weak.

Score:0

Crypto

Titanlord

1/20/24, 11:32 AM

I'm not an expert on this field, but because there is no answer yet, I give it a try.

My university had some projects on data privacy regarding machine learning and data privacy for medical data (are those enough identifiers to identify my university? :D). I would say the projects for medical data belong to big data and the main approaches were k-anonymity (because of simplicity) and differential privacy.

I would count machine learning to AI and those projects tried to secure the privacy of the data on which the networks learn on. As far as I understand it, the same techniques were used to secure the privacy of the input data for learning.

To simplify the ideas: You can assume, that machine learning uses big data to train networks. If the original data is not secure/privat, the resulting network is neither. And vice versa they stated security/privacy of the network (of course with a much more complex argumentation).

But AI and Big Data are buzzwords covering a lot of topics and those projects are only for specific applications (mainly medical). Therefore this may not be correct for different AI-approaches, or even different use cases of machine learning. Nevertheless, I hope that I have been able to provide a little food for thought.

+ 0

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: General Questions on Big Data and AI privacy

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.