Key management problem in federated learning based on homomorphic encryption

sunmu

4/17/24, 2:12 AM

In federated learning using homomorphic encryption, all participants in most schemes share the same pair of keys, which can easily cause key leaks and lead to data privacy leaks.

After research, I found that someone proposed to use a multi-key homomorphic encryption algorithm to solve the above problems. But I wonder why no one has added a key management scheme, such as secret sharing, to the existing single-key homomorphic encryption schemes, so that each participant has only a part of the key.

This can also solve the problem of single-key encryption in federated learning. The key homomorphic encryption key is easy to leak. Is it feasible to apply secret sharing as a key management scheme in federated learning based on homomorphic encryption?

1 + 0

homomorphic-encryption

machine-learning

Score:1

Crypto

vince.h

4/17/24, 9:03 AM

Indeed people have studied secret sharing + homomorphic encryption, it's actually one of those multi-party homomorphic encryption schemes. Check https://eprint.iacr.org/2021/1412 for some introduction.

+ 0

Elon Musk

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

EN: Key management problem in federated learning based on homomorphic encryption

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.