Laplacian ISI Eigenvalue Based Anomaly Detection in IoT Networks

Abstract

Anomaly detection in Internet of Things networks is a critical challenge due to dynamic topol ogy, heterogeneous devices, and complex communi cation patterns. This paper presents a novel spec tral framework based on the Laplacian Inverse Sum In-degree matrix for detecting anomalies in graph structured data. The proposed approach models IoT systems as graphs and extracts Laplacian ISI eigenvalues as discriminative features that capture both structural connectivity and degree-based re lationships. Unlike traditional Laplacian methods, the ISI formulation provides enhanced sensitivity to structural perturbations. The extracted spectral fea tures are integrated with machine learning models for anomaly classification. Experimental analysis on simulated network data demonstrates that devia tions in the Laplacian ISI spectrum effectively iden tify abnormal behavior such as node failures and ma licious activities. The proposed framework bridges spectral graph theory and practical anomaly detec tion, making it suitable for cybersecurity and IoT applications.