PoCoFL: POlicy-COmpliant Federated Learning
PoCoFL uses commitments and non-interactive zero-knowledge proofs to verify policy-compliant client contributions and correct aggregation in federated learning.
PoCoFL is a policy-compliant federated learning framework that decouples FL type, policy semantics, and cryptographic realization, making it network-topology agnostic. Clients prove contribution compliance via commitments and NIZK proofs while aggregators prove admitted contributions were processed per the selected policy. The authors formalize four instantiations: vanilla, continual, personalised, and threshold-encrypted FL, and provide proof-of-concept implementations. Evaluations measure the effect of policy enforcement on learning objectives across vanilla, personalised, and continual FL.
- Decouples FL type, policy semantics, and cryptographic construction for topology-agnostic validation
- Clients prove compliance with commitments and NIZK proofs; aggregators prove correct processing
- Covers vanilla, continual, personalised, and threshold-encrypted FL instantiations
- Proof-of-concept implementations demonstrate practical feasibility
From the original article
1 min read at arxiv.orgFederated Learning (FL) is a privacy-oriented learning paradigm that enables collaborative model training while keeping training data local to participating clients. However, it does not guarantee that clients submit policy-compliant contributions or that aggregators process admitted contributions correctly.
By Dominik Roy George · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.