Even with partially homomorphic encryption and additional "privacy preserving protocols" you can carry out pretty general computation tasks such as machine lear

Even with partially homomorphic encryption and additional "privacy preserving protocols" you can carry out pretty general computation tasks such as machine learning. Have a look at this blog post using the Paillier cryptosystem for a federated linear regression - https://blog.n1analytics.com/distributed-machine-learning-and-partially-homomorphic-encryption-1/ https://blog.n1analytics.com/distributed-machine-learning-an...