Detecting abnormal nodes in IoT security using neural networks and graph theory

Articles in Press

Document Type : Original Article

Authors

1 department of computer engineering, NT.C. Islamic Azad University

2 Department of Computer Engineering, SR.C., Islamic Azad University, Tehran, Iran

Abstract

The increasing use of Internet of Things (IoT) devices has introduced new security challenges, making timely detection of anomalies and abnormal nodes essential. Limitations in hardware resources, protocol diversity, and network traffic complexity render traditional methods ineffective in identifying threats. This paper presents a hybrid approach based on graph theory and artificial neural networks to classify network nodes into three categories: healthy, suspicious, and unhealthy. Graph features such as centrality measures are extracted, and anomaly detection accuracy is improved to 99.03% using LOWESS curve fitting, compared to 98.22% for a neural network without curve fitting and 98.87% for linear and polynomial models. This method outperforms other approaches and serves as an effective tool for improving security in IoT networks.

Keywords

Main Subjects

Tabriz Journal of Electrical Engineering

Articles in Press, Corrected Proof
Available Online from 06 May 2026

Files

Share

How to cite

Statistics