Detecting Cyber Attacks using the Matrix Profile

Abstract

Cyber-attacks have been increasing in recent years and are becoming a threat to IoT infrastructures. Anomaly Detection can detect the attacks earlier and form an early warning system. In this work, the overall goal is to detect Denial of Service (DoS) attacks aimed against a Smart Farming Infrastructure. We are using the Matrix Profile (MP) for anomaly detection on network traffic data that has been collected during a deauthentication attack. The MP is a data structure developed by Eamonn Keogh for time series analysis. While past work has used machine learning anomaly detection approaches such as Autoencoder on network traffic, no other work has applied the MP on network traffic data for anomaly detection of cyber-attacks. Since the dataset is labeled, we will evaluate our method using standard anomaly detection metrics including accuracy, precision, F1, and recall.