Papers · Federated & aggregation · RoFL

RoFL

RoFL: Attestable Robustness for Secure Federated Learning

Lukas Burkhalter, Hidde Lycklama, Alexander Viand, Nicolas Küchler, Anwar Hithnawi

Venue
IEEE S&P 2023
Date
2021-07-07
Uses ZK
yes
Paper
https://arxiv.org/abs/2107.03311
Code
https://github.com/pps-lab/rofl-project-code

Notes

Same group as Artemis (Lycklama, Viand, Küchler, Hithnawi). Requires a commitment per vector entry, so communication overhead is substantial -- the recurring cost of ZK-over-secure-aggregation.

Our reading

Citation neighbourhood

Builds on
  • none recorded
Cited by, in this corpus

Edges are a proxy: paper A's text mentions B anywhere (body or bibliography). See the full graph.

Discussed in

The hybrid cellWhat each system actually provesEvery relaxation, and what an adversary can still do
Other recorded fields
primitive: Zero-knowledge proofs (norm bounds, e.g. L2/Linf) over commitments to encrypted
  updates
what_is_proven: Each client's update satisfies declared constraints -- NOT that the
  update came from correct training.