A 1D CNN-based model for IoT anomaly detection using INT data

Abstract

Due to the limited capacity and versatility of Internet of Things (IoT) devices, it isn't easy to implement advanced security mechanisms and adhere to common security standards on IoT devices. Our study proposes a network-based solution to address these issues in the IoT environment. This solution leverages the advantages of a programmable data plane, Software-Defined Networking (SDN), and machine learning. In-Band Network Telemetry (INT) is a novel monitoring application developed using a programmable data plane to collect network characteristics (INT data) in real time without affecting network performance. We aim to detect IoT attacks based on INT data using a 1D CNN-based deep learning model. As far as we know, this model is the first attempt to use INT data to detect IoT attacks. We created an SDN network infrastructure in a simulation environment and collected INT data from IoT devices in the event of an attack or non-attack. Our proposed 1D CNN-based model using INT data can detect IoT attacks with approximately 99.63 % accuracy. Our solution is relatively cost-effective and performs well compared to other competing models.