Ensemble Learning-based Sybil Attacks Detection in Smart Meter Network

A smart meter provides real-time monitoring and recording of the electricity, gas, or water at a consumer level. Further, smart meters offer remote monitoring, better resource management, and the ability to provide detailed usage information to both consumers and utility companies. However, smart meters are prone to significant potential threats, vulnerabilities, and risks that could compromise the security and functionality of smart meters as well as the privacy and security of the consumer. A Sybil attack is one such malicious activity that creates a multiple number of fake nodes, accounts, or identities to deceive the system or network into thinking that these fake identities are real and legitimate. Thus, several security measures are required to address these security challenges by Sybil Attack including digital signatures and authentication, anomaly detection, secure communication, regular auditing, redundancy checks, and Machine Learning (ML)-based approaches. Amongst these measures, ML-based algorithms play a significant role in the real-time detection of Sybil attacks by anomaly detection, behavior profiling, and classification. In this paper, we use several state-of-the-art ML classifiers for real-time detection of Sybil attacks and anomalies in user behavior or network activity. Further, we use an Ensemble Learning method that combines the predictions of multiple classifiers to enhance accuracy and reduce false positives. The Ensemble Learning-based classifier outperforms the individual classifiers in the detection of Sybil attacks.