Identification of strategic sensor locations for intrusion detection and classification in smart grid networks

Abstract

The dependence of smart grid (SG) on advanced communication technology has increased its vulnerability to both cyber and physical attacks. Unlike, a physical attack which can be easily detected by the state estimation (SE) task performed at control center, cyberattack detection is accomplished using the information obtained through the entire set of secured sensors. However, the economic constraints hinder securing the entire monitoring system. In this regard, the present work proposes a cost-effective framework for detecting physical, cyber and cyber–physical(CP) attacks by imparting security to a minimal set of sensors. The attack detection scheme involves identifying a minimal sensor set using a graph theory-based approach for topological observability. The attack detection task is executed by solving AC state estimation (AC-SE) for varying types of attacks based on the secured sensor information at strategic locations. Prioritizing strategic locations allows for maximizing detection accuracy under different levels of budget for security. The formulation does not assume any restriction on the accessibility of sensor information to the intruder to launch an attack. Following attack detection, the attack type is determined based on sensor measurements. The scheme has been extensively validated for different systems in terms of attack detection and classification accuracy under different levels of budget availability and intruder accessibility.