IoT Anomaly Detection with 1D CNN Using P4 Capabilities

Gereltsetseg Altangerel, Máté Tejfel, Enkhtur Tsogbaatar

IoT Anomaly Detection with 1D CNN Using P4 Capabilities

Číslo: 2/2023
Periodikum: Acta Electrotechnica et Informatica
DOI: 10.2478/aei-2023-0006

Klíčová slova: P4, In-Band Network Telemetry(INT), IoT security, 1D-CNN, SDN

Pro získání musíte mít účet v Citace PRO.

Přečíst po přihlášení

Anotace: Although the Internet of Things (IoT) is a rapidly developing technology, it also brings a number of security challenges, such asIoT attacks. Currently, research on IoT anomaly detection in Software-Defined Networking (SDN) relies only on the control plane.In this study, we aim to detect IoT anomalies by covering the advantages of the control and data plane. First, we collected real-timenetwork telemetry data from the data plane based on the capabilities of the P4. Then, using this telemetry data, we built differentanomaly detection models and compared their performance. Among them, the one-Dimensional Convolutional Neural Network (1DCNN) model classified our data best and showed the highest performance, so we proposed this model for IoT anomaly detection on thecontrol plane. To our knowledge, our approach is the first solution that integrates the control plane and data plane for IoT anomalydetection. Finally, when evaluating the performance of our proposed 1D CNN model, the accuracy, F1 score, and Matthews correlationcoefficient (MCC) are the same or better than existing studies