Hasty Atashzar - Iran Energy Efficiency Organization(SABA) Tehran, Iran
Hadi Modaghegh - Iran Energy Efficiency Organization(SABA)Tehran, Iran

Advanced Metering Infrastructures have great potential to improve the stability and reliability of the electric power grid and will become key tools to empower consumers in the energy market. However, if they are not based upon a secure system architecture, they could in fact become one of the grid’smost significant liabilities, due to their expected pervasive deployment. One of the most significant cyber attacks which threaten the availability of AMI components is DDoS attack. This paper concentrates on applying machine learning techniques to defence against DDOS attacks in AMI system. Accordingly, an effective classifier, PCNN, is proposed. PCNN classifier consists of two major parts: Principal Component Analysis (PCA) for feature extraction and MLP Neural Network (NN) for packet classification. Experimental dataset generated by Spirnet Avalanche in our Lab has been employed to evaluate and examine the proposed classifier. Results demonstrate that PCNN classifier is able to classify the incoming traffic with less than 7% false acceptance rate and less than 4% false rejection rate at the edge of the victim network

AMI Security; Traffic classification; Machine learning; Principal Component Analysis; Artificial Neural Network