Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficient Personalized Federated PCA with Manifold Optimization for IoT Anomaly Detection

Xianchao Xiu, Chenyi Huang, Wei Zhang, Wanquan Liu

TL;DR

A manifold optimization algorithm based on the alternating direction method of multipliers (ADMM) with rigorous theoretical convergence guarantees is developed, which outperforms the state-of-the-art FedPG, achieving excellent F1-scores and accuracy in various IoT security scenarios.

Abstract

Internet of things (IoT) networks face increasing security threats due to their distributed nature and resource constraints. Although federated learning (FL) has gained prominence as a privacy-preserving framework for distributed IoT environments, current federated principal component analysis (PCA) methods lack the integration of personalization and robustness, which are critical for effective anomaly detection. To address these limitations, we propose an efficient personalized federated PCA (FedEP) method for anomaly detection in IoT networks. The proposed model achieves personalization through introducing local representations with the $\ell_1$-norm for element-wise sparsity, while maintaining robustness via enforcing local models with the $\ell_{2,1}$-norm for row-wise sparsity. To solve this non-convex problem, we develop a manifold optimization algorithm based on the alternating direction method of multipliers (ADMM) with rigorous theoretical convergence guarantees. Experimental results confirm that the proposed FedEP outperforms the state-of-the-art FedPG, achieving excellent F1-scores and accuracy in various IoT security scenarios. Our code will be available at \href{https://github.com/xianchaoxiu/FedEP}{https://github.com/xianchaoxiu/FedEP}.

Efficient Personalized Federated PCA with Manifold Optimization for IoT Anomaly Detection

TL;DR

A manifold optimization algorithm based on the alternating direction method of multipliers (ADMM) with rigorous theoretical convergence guarantees is developed, which outperforms the state-of-the-art FedPG, achieving excellent F1-scores and accuracy in various IoT security scenarios.

Abstract

Internet of things (IoT) networks face increasing security threats due to their distributed nature and resource constraints. Although federated learning (FL) has gained prominence as a privacy-preserving framework for distributed IoT environments, current federated principal component analysis (PCA) methods lack the integration of personalization and robustness, which are critical for effective anomaly detection. To address these limitations, we propose an efficient personalized federated PCA (FedEP) method for anomaly detection in IoT networks. The proposed model achieves personalization through introducing local representations with the -norm for element-wise sparsity, while maintaining robustness via enforcing local models with the -norm for row-wise sparsity. To solve this non-convex problem, we develop a manifold optimization algorithm based on the alternating direction method of multipliers (ADMM) with rigorous theoretical convergence guarantees. Experimental results confirm that the proposed FedEP outperforms the state-of-the-art FedPG, achieving excellent F1-scores and accuracy in various IoT security scenarios. Our code will be available at \href{https://github.com/xianchaoxiu/FedEP}{https://github.com/xianchaoxiu/FedEP}.
Paper Structure (27 sections, 1 theorem, 30 equations, 6 figures, 8 tables)

This paper contains 27 sections, 1 theorem, 30 equations, 6 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Notations
  4. Sparse PCA
  5. FedPCA for IoT Anomaly Detection
  6. The Proposed Method
  7. Problem Formulation
  8. Optimization Algorithm
  9. Update $W_i$
  10. Update Si
  11. Update Ui
  12. Update V
  13. Update Lambda-i and Pi-i
  14. Convergence Analysis
  15. Experiments
  16. ...and 12 more sections

Key Result

Theorem 3.1

The augmented Lagrangian function sequence $\{\mathcal{L} (\{W_i^k\}, \{S_i^k\}, \{U_i^k\}, V^k; \{\Lambda_i^k\}, \{\Pi_i^k\})\}$ is nonincreasing.

Figures (6)

  • Figure 1: Framework of the proposed FedEP.
  • Figure 2: ROC curves on (a) TON-IoT, (b) UNSW-NB15, (c) NSL-KDD.
  • Figure 3: Accuracy over communication rounds on (a) TON-IoT, (b) UNSW-NB15, (c) NSL-KDD.
  • Figure 4: Feature importance visualization on (a) TON-IoT, (b) UNSW-NB15, (c) NSL-KDD.
  • Figure 5: Training time per round as the number of clients scales on (a) TON-IoT, (b) UNSW-NB15, (c) NSL-KDD.
  • ...and 1 more figures

Theorems & Definitions (2)

  • Theorem 3.1
  • Remark 3.2