Federated Learning is a way to train a machine learning model on user data distributed across many devices in a way that supposedly allows the data to stay private. In this blog post, we give an overview of the technology and its applications and show how these privacy claims can get problematic very quickly. Any feedback is welcome!
At Pyte, we don't use HE, but rather a related but much more practical primitive (Secure Multi-Party Computation, SMPC). However, this survey post meant to be merely educational, since HE is a very fascinating paradigm (but I agree, in the present form it is impractical and will be for the foreseeable future).
Homomorphic Encryption is a modern way of encryption that allows one to run algorithms on encrypted data without first decrypting it. We give an overview of the technology, discuss its many flavors and compare it with related privacy-enhancing technologies. Any feedback is welcome!