Microsecond Federated SVD on Grassmann Manifold for Real-time IoT Intrusion Detection

TL;DR

FedSVD is an innovative federated learning framework that uses Grassmann manifold optimization and SVD for real-time, unsupervised intrusion detection in IoT networks, reducing communication and computation costs.

Contribution

It introduces FedSVD, a novel unsupervised federated learning approach utilizing Grassmann manifold optimization for efficient IoT intrusion detection.

Findings

Achieves comparable accuracy to deep learning methods.

Reduces inference latency by over 10 times.

Operates efficiently on low-power IoT devices.

Abstract

This paper introduces FedSVD, a novel unsupervised federated learning framework for real-time anomaly detection in IoT networks. By leveraging Singular Value Decomposition (SVD) and optimization on the Grassmann manifolds, FedSVD enables accurate detection of both known and unknown intrusions without relying on labeled data or centralized data sharing. Tailored for deployment on low-power devices like the NVIDIA Jetson AGX Orin, the proposed method significantly reduces communication overhead and computational cost. Experimental results show that FedSVD achieves performance comparable to deep learning baselines while reducing inference latency by over 10x, making it suitable for latency-sensitive IoT applications.