Enhancing the cyber-security of smart grids with applications to synchrophasor data

Abstract

In the power grids, Supervisory Control and Data Acquisition (SCADA) systems are used as part of the Energy Management System (EMS) for enabling grid monitoring, control and protection. In recent times, with the ongoing installation of thousands of Phasor Measurement Units (PMUs), system operators are becoming increasingly reliant on PMU-generated synchrophasor measurements for executing wide-area monitoring and real-time control. The availability of PMU data facilitates dynamic state estimation of the system, thus improving the efficiency and resiliency of the grid. Since the SCADA and PMU data are used to make critical control decisions including actuation of physical systems, the timely availability and integrity of this networked data is of paramount importance. Absence or wrong control actions can potentially lead to disruption of operations, monetary loss, damage to equipments or surroundings or even blackout. This has posed new challenges to information security especially in this age of ever-increasing cyber-attacks. In this thesis, potential cyber-attacks on smart grids are presented and effective and implementable schemes are proposed for detecting them. The focus is mainly on three kinds of cyber-attacks and their detection: (i) gray-hole attacks on synchrophasor systems, (ii) PMU data manipulation attacks and (iii) data integrity attacks on SCADA systems.

The proposed detection mechanisms have demonstrated high accuracy in real-time detection of attacks of various magnitudes, simulated on real PMU data obtained from the NY grid. By performing alarm clustering, the occurrence of false alarms has been reduced to almost zero. The solutions are computationally inexpensive, low on cost, do not add any overhead, and do not require any feedback from the network.

The scheme for detecting data integrity attacks on SCADA systems is based on utilizing synchrophasor measurements from available PMUs in the grid. The proposed method uses a difference measure, developed in this thesis, to determine the relative divergence and mis-correlation between the datasets. Based on the estimated difference measure, tampered and genuine data can be distinguished.

In the case of PMU data manipulation attacks, the attacker may modify the data in the PMU packets in order to bias the system states and influence the control center into taking wrong decisions. The proposed detection technique is based on evaluating the equivalent impedances of the transmission lines and classifying the observed anomalies to determine the presence of attack and its location.

In the case of gray-hole attacks, also known as packet-drop attacks, the adversary may arbitrarily drop PMU data packets as they traverse the network, resulting in unavailability of time-sensitive data for the various critical power system applications. The fundamental challenge is to distinguish packets dropped by the adversary from those that occur naturally due to network congestion.The proposed gray-hole attack detection technique is based on exploiting the inherent timing information in the GPS time-stamped PMU data packets and using the temporal trends of the latencies to classify the cause of packet-drops and finally detect attacks, if any.