A Privacy Enhancing Federated Learning Framework for Multi-Party Distributed ..
summary
Claim: keep raw data local and train global models instead of moving terabytes of energy data. Built a proof‑of‑concept federated learning platform using Flower that trains next‑4‑hour household prediction local models on real heat‑pump measurements and aggregates to a global model and simulator (ESL standard). Finds feasibility before anonymization but flags attacks in low‑voltage zones with ~12 houses, shows homomorphic encryption bloats data and recommends mixed anonymization, synthetic data and differential‑privacy engineering rather than naive encryption.
AI-written summary · about
chapters (8)
- 0:00Federated learning basics
- 4:07Keeping data local
- 7:41Vertical FL challenges
- 10:16Energy sector FL cases
- 14:58Building FL with Flower
- 17:56Local predictions for energy
- 21:58Privacy anonymization
- 26:27European data spaces
related talks
26mMachine Learning without the Data #federatedlearning
29mBalancing Thin Line Between Data Intelligence & Privacy
27mCreate Value and Insight Through Privacy-preserving Data Collaboration [...]
32mInnovation in the Age of Regulation: Federated Learning with Flower
42mLearning Together, Distributed: An Introduction to Federated Learning
25mData collaboration with Federated Learning
41mFederated Learning: Machine Learning on Decentralized Data
40mDemystifying Decentralized Fair Models for Federated Machine Learning