Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FedNC: A Secure and Efficient Federated Learning Method with Network Coding

Yuchen Shi, Zheqi Zhu, Pingyi Fan, Khaled B. Letaief, Chenghui Peng

TL;DR

This paper introduces FedNC, an NC-inspired Federated Learning framework that encodes local model updates using Random Linear Network Coding before uploading for aggregation. By transmitting encoded packets $C_i=\boldsymbol{a_i}P=\sum_{k=1}^K \alpha_{ik} w_k^{(t)}$ over a finite field, FedNC enables server-side decoding whenever the coding matrix $A$ is invertible, improving security, throughput, robustness, and efficiency. The authors derive an upper bound on the transmission error probability $p_e\le 1-\left(1-2^{-s}\right)^{\eta}$ and validate the approach on CIFAR-10 with iid and mixed non-iid data, showing significant gains in robustness and convergence speed, especially in large-scale settings. The work suggests that network coding concepts can provide practical, scalable improvements for secure and efficient FL, inviting future convergence analysis and richer evaluations.

Abstract

Federated Learning (FL) is a promising distributed learning mechanism which still faces two major challenges, namely privacy breaches and system efficiency. In this work, we reconceptualize the FL system from the perspective of network information theory, and formulate an original FL communication framework, FedNC, which is inspired by Network Coding (NC). The main idea of FedNC is mixing the information of the local models by making random linear combinations of the original parameters, before uploading for further aggregation. Due to the benefits of the coding scheme, both theoretical and experimental analysis indicate that FedNC improves the performance of traditional FL in several important ways, including security, efficiency, and robustness. To the best of our knowledge, this is the first framework where NC is introduced in FL. As FL continues to evolve within practical network frameworks, more variants can be further designed based on FedNC.

FedNC: A Secure and Efficient Federated Learning Method with Network Coding

TL;DR

This paper introduces FedNC, an NC-inspired Federated Learning framework that encodes local model updates using Random Linear Network Coding before uploading for aggregation. By transmitting encoded packets over a finite field, FedNC enables server-side decoding whenever the coding matrix is invertible, improving security, throughput, robustness, and efficiency. The authors derive an upper bound on the transmission error probability and validate the approach on CIFAR-10 with iid and mixed non-iid data, showing significant gains in robustness and convergence speed, especially in large-scale settings. The work suggests that network coding concepts can provide practical, scalable improvements for secure and efficient FL, inviting future convergence analysis and richer evaluations.

Abstract

Federated Learning (FL) is a promising distributed learning mechanism which still faces two major challenges, namely privacy breaches and system efficiency. In this work, we reconceptualize the FL system from the perspective of network information theory, and formulate an original FL communication framework, FedNC, which is inspired by Network Coding (NC). The main idea of FedNC is mixing the information of the local models by making random linear combinations of the original parameters, before uploading for further aggregation. Due to the benefits of the coding scheme, both theoretical and experimental analysis indicate that FedNC improves the performance of traditional FL in several important ways, including security, efficiency, and robustness. To the best of our knowledge, this is the first framework where NC is introduced in FL. As FL continues to evolve within practical network frameworks, more variants can be further designed based on FedNC.
Paper Structure (20 sections, 2 theorems, 11 equations, 4 figures, 1 table, 1 algorithm)

This paper contains 20 sections, 2 theorems, 11 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works and Motivations
  3. Contributions and Paper Organization
  4. Preliminaries
  5. Federated Learning Problem
  6. Random Linear Network Coding
  7. FedNC: Framework and Algorithm
  8. Benefits of FedNC
  9. Security
  10. Troughput Gain
  11. Robustness
  12. Efficiency
  13. Error probability of transmission in FedNC
  14. Numerical Results
  15. Experiment Setting
  16. ...and 5 more sections

Key Result

Proposition 1

Consider an FL system and assume that the central server would like to collect the packets from all clients in the set $\mathcal{P}_t$, where $|\mathcal{P}_t|=K$, and the number of packets for each client is the same. Assume that the server selects packets according to random sampling. Let $G$ be th where $\gamma\approx 0.577$ is the Euler–Mascheroni constant.

Figures (4)

  • Figure 1: The schematic of the coding process in LNC.
  • Figure 2: Traditional FL and FedNC structures. Although both are vulnerable to attack threats during transmission over open channels, FedNC transmits encoded packets while traditional FL transmits original packets.
  • Figure 3: Comparisons between classical FL ($\mathtt{FedAvg}$) and FedNC with numerous settings of field size $s$ and number of links $\eta$ under two different data splitting, where $N=100$.
  • Figure 4: Comparisons between classical FL ($\mathtt{FedAvg}$) and FedNC ($s=1, \eta=8$) under different system scales.

Theorems & Definitions (5)

  • Proposition 1
  • proof
  • Remark 1
  • Proposition 2
  • proof