Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Standing Firm in 5G: A Single-Round, Dropout-Resilient Secure Aggregation for Federated Learning

Yiwei Zhang, Rouzbeh Behnia, Imtiaz Karim, Attila A. Yavuz, Elisa Bertino

TL;DR

The paper tackles privacy-preserving federated learning in 5G, where high device churn and dropouts hinder secure aggregation. It introduces a single-round secure aggregation protocol that leverages base-station assisted unmasking, precomputation, key-homomorphic pseudorandom functions (KHPRFs), and $t$-out-of-$k$ secret sharing to tolerate dropouts and reduce overhead. A thorough security analysis under a realistic threat model is complemented by experimental results showing lower computation and communication costs, along with robust resilience to UE and BS dropouts. The approach enables scalable, private FL deployments in real-world 5G networks, with potential impact on edge analytics and network optimization.

Abstract

Federated learning (FL) is well-suited to 5G networks, where many mobile devices generate sensitive edge data. Secure aggregation protocols enhance privacy in FL by ensuring that individual user updates reveal no information about the underlying client data. However, the dynamic and large-scale nature of 5G-marked by high mobility and frequent dropouts-poses significant challenges to the effective adoption of these protocols. Existing protocols often require multi-round communication or rely on fixed infrastructure, limiting their practicality. We propose a lightweight, single-round secure aggregation protocol designed for 5G environments. By leveraging base stations for assisted computation and incorporating precomputation, key-homomorphic pseudorandom functions, and t-out-of-k secret sharing, our protocol ensures efficiency, robustness, and privacy. Experiments show strong security guarantees and significant gains in communication and computation efficiency, making the approach well-suited for real-world 5G FL deployments.

Standing Firm in 5G: A Single-Round, Dropout-Resilient Secure Aggregation for Federated Learning

TL;DR

The paper tackles privacy-preserving federated learning in 5G, where high device churn and dropouts hinder secure aggregation. It introduces a single-round secure aggregation protocol that leverages base-station assisted unmasking, precomputation, key-homomorphic pseudorandom functions (KHPRFs), and -out-of- secret sharing to tolerate dropouts and reduce overhead. A thorough security analysis under a realistic threat model is complemented by experimental results showing lower computation and communication costs, along with robust resilience to UE and BS dropouts. The approach enables scalable, private FL deployments in real-world 5G networks, with potential impact on edge analytics and network optimization.

Abstract

Federated learning (FL) is well-suited to 5G networks, where many mobile devices generate sensitive edge data. Secure aggregation protocols enhance privacy in FL by ensuring that individual user updates reveal no information about the underlying client data. However, the dynamic and large-scale nature of 5G-marked by high mobility and frequent dropouts-poses significant challenges to the effective adoption of these protocols. Existing protocols often require multi-round communication or rely on fixed infrastructure, limiting their practicality. We propose a lightweight, single-round secure aggregation protocol designed for 5G environments. By leveraging base stations for assisted computation and incorporating precomputation, key-homomorphic pseudorandom functions, and t-out-of-k secret sharing, our protocol ensures efficiency, robustness, and privacy. Experiments show strong security guarantees and significant gains in communication and computation efficiency, making the approach well-suited for real-world 5G FL deployments.
Paper Structure (15 sections, 1 theorem, 1 equation, 4 figures)

This paper contains 15 sections, 1 theorem, 1 equation, 4 figures.

Table of Contents

  1. Introduction
  2. Preliminaries and Models
  3. System Model
  4. Threat Model
  5. Our Protocol
  6. Setup Phase
  7. Aggregation Phase
  8. Security Analysis
  9. Evaluation
  10. Experiment Setup
  11. Results
  12. Efficiency
  13. Resilience to UE and BS Dropout
  14. Summary of Findings
  15. Conclusion and Future Work

Key Result

theorem 1

The protocol proposed above with $n$ users, $k$ base stations and the aggregation server $\mathsf{AF_{FL}}$ is private against an adversary $\mathcal{A}$ where $\mathcal{A}$ is able to compromise $(1-\alpha)n$ users, $k-1$ base stations and the aggregation server $\mathsf{AF_{FL}}$.

Figures (4)

  • Figure 1: 5G+FL Framework.
  • Figure 2: 5G-based FL secure aggregation protocol.
  • Figure 3: Efficiency Comparison.
  • Figure 4: Dropout Resilience.

Theorems & Definitions (5)

  • definition 1: Key-Homomorphic Pseudorandom Function (KHPRF) boneh2013key
  • definition 2: $t$-out-of-$n$ Secret Sharing shamir1979share
  • definition 3: $\alpha$-summation ideal functionality bell2020secure
  • definition 4: Privacy of Secure Aggregation bell2020secure
  • theorem 1