Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

VerifBFL: Leveraging zk-SNARKs for A Verifiable Blockchained Federated Learning

Ahmed Ayoub Bellachia, Mouhamed Amine Bouchiha, Yacine Ghamri-Doudane, Mourad Rabah

TL;DR

VerifBFL tackles the challenge of achieving end-to-end verifiability and privacy in blockchain-based federated learning for crowdsourcing. It combines Nova recursive zk-SNARKs with Incrementally Verifiable Computation to produce proofs of local model accuracy and global aggregation, which are verified on-chain, while differential privacy protects training data. The architecture leverages IPFS for off-chain storage, a PBFT permissioned blockchain for governance, and Chainlink DON to offload heavy verification tasks, delivering practical proof-generation and verification times (e.g., training proofs <81s, aggregation proofs <2s, on-chain verification <0.6s). Experimental results demonstrate feasibility within a crowdsourcing setting, though the authors acknowledge scalability as an area for further improvement and optimization under higher loads.

Abstract

Blockchain-based Federated Learning (FL) is an emerging decentralized machine learning paradigm that enables model training without relying on a central server. Although some BFL frameworks are considered privacy-preserving, they are still vulnerable to various attacks, including inference and model poisoning. Additionally, most of these solutions employ strong trust assumptions among all participating entities or introduce incentive mechanisms to encourage collaboration, making them susceptible to multiple security flaws. This work presents VerifBFL, a trustless, privacy-preserving, and verifiable federated learning framework that integrates blockchain technology and cryptographic protocols. By employing zero-knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARKs) and incrementally verifiable computation (IVC), VerifBFL ensures the verifiability of both local training and aggregation processes. The proofs of training and aggregation are verified on-chain, guaranteeing the integrity and auditability of each participant's contributions. To protect training data from inference attacks, VerifBFL leverages differential privacy. Finally, to demonstrate the efficiency of the proposed protocols, we built a proof of concept using emerging tools. The results show that generating proofs for local training and aggregation in VerifBFL takes less than 81s and 2s, respectively, while verifying them on-chain takes less than 0.6s.

VerifBFL: Leveraging zk-SNARKs for A Verifiable Blockchained Federated Learning

TL;DR

VerifBFL tackles the challenge of achieving end-to-end verifiability and privacy in blockchain-based federated learning for crowdsourcing. It combines Nova recursive zk-SNARKs with Incrementally Verifiable Computation to produce proofs of local model accuracy and global aggregation, which are verified on-chain, while differential privacy protects training data. The architecture leverages IPFS for off-chain storage, a PBFT permissioned blockchain for governance, and Chainlink DON to offload heavy verification tasks, delivering practical proof-generation and verification times (e.g., training proofs <81s, aggregation proofs <2s, on-chain verification <0.6s). Experimental results demonstrate feasibility within a crowdsourcing setting, though the authors acknowledge scalability as an area for further improvement and optimization under higher loads.

Abstract

Blockchain-based Federated Learning (FL) is an emerging decentralized machine learning paradigm that enables model training without relying on a central server. Although some BFL frameworks are considered privacy-preserving, they are still vulnerable to various attacks, including inference and model poisoning. Additionally, most of these solutions employ strong trust assumptions among all participating entities or introduce incentive mechanisms to encourage collaboration, making them susceptible to multiple security flaws. This work presents VerifBFL, a trustless, privacy-preserving, and verifiable federated learning framework that integrates blockchain technology and cryptographic protocols. By employing zero-knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARKs) and incrementally verifiable computation (IVC), VerifBFL ensures the verifiability of both local training and aggregation processes. The proofs of training and aggregation are verified on-chain, guaranteeing the integrity and auditability of each participant's contributions. To protect training data from inference attacks, VerifBFL leverages differential privacy. Finally, to demonstrate the efficiency of the proposed protocols, we built a proof of concept using emerging tools. The results show that generating proofs for local training and aggregation in VerifBFL takes less than 81s and 2s, respectively, while verifying them on-chain takes less than 0.6s.
Paper Structure (27 sections, 3 figures, 2 tables)

This paper contains 27 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Incrementally Verifiable Computation
  4. Nova zk-SNARK
  5. Differential Privacy
  6. Related work
  7. System Design
  8. System Model
  9. Threat Model
  10. VerifBFL Arichitecture
  11. IPFS
  12. Blockchain
  13. Decentralized Oracle Network
  14. Workflow
  15. zk-SNARK Proofs Construction
  16. ...and 12 more sections

Figures (3)

  • Figure 1: Design Overview: VerifBFL is composed of two layers, with the blockchain layer overseeing the work of the FL layer. TPs, TAs, and AGs store FL tasks metadata on IPFS, then submit Txs to the blockchain in a crowdsourced manner. Proofs for training and aggregation are generated off-chain by TAs and AGs and verified on-chain.
  • Figure 2: Chainlink Request Model: the consumer contract sends requests to Chainlink nodes through the operator contract, custom computations are performed by leveraging external adapters.
  • Figure 4: Latency and Throughput of VerifBFL

Theorems & Definitions (1)

  • Definition 2.1: $\epsilon$-Differential Privacy