Lightning Talk: Multi-Party Learning Using Homomorphic Encryption with Kubeflow...
summary
Describes an implemented workflow that uses Kubeflow Pipelines orchestration to preprocess private datasets locally, encrypt them with homomorphic encryption on encrypted data, and run model fitting and inference without exposing raw inputs. Demonstration uses small datasets and reports that HE computation proves very expensive in practice, limiting scale. Claims next steps include collaboration with Boston University on hardware accelerators to reduce cost. Lists concrete target domains: healthcare, electric grids, transportation, and government.
AI-written summary · about
related talks
52mMachine Learning over Encrypted Data with Fully Homomorphic Encryption
25mData collaboration with Federated Learning
28mA Privacy Enhancing Federated Learning Framework for Multi-Party Distributed ..
14mLightning Talk: Leveraging Kubeflow for Next-Generation AI Healthcare Solutions
37mOpen-source Machine Learning on Encrypted Data
18mKeynote: Containerizing Deep Learning Using Kubeflow on AWS
40mTrust No One
44mIP Protection and Privacy in LLM: Leveraging Fully Homomorphic Encryption