An Anomaly-Based Intrusion Detection System for the Smart Grid Based on CART Decision Tree

Abstract

The Smart Grid (SG) paradigm constitutes the new technological evolution of the traditional electrical grid, providing remote monitoring and controlling capabilities among all its operations through computing services. These new capabilities offer a lot of benefits, such as better energy management, increased reliability and security, as well as more economical pricing. However, despite these advantages, it introduces significant security challenges, as the computing systems and the corresponding communications are characterized by several cybersecurity threats. An efficient solution against cyberattacks is the Intrusion Detection Systems (IDS). These systems usually operate as a second line of defence and have the ability to detect or even prevent cyberattacks in near real-time. In this paper, we present a new IDS for the Advanced Metering Infrastructure (AMI) utilizing machine learning capabilities based on a decision tree. Decision trees have been used for multiple classification problems like the distinguishment between the normal and malicious activities. The experimental evaluation demonstrates the efficiency of the proposed IDS, as the Accuracy and the True Positive Rate of our IDS reach 0.996 and 0.993 respectively.