Society is increasingly dependent on the reliable operation of power systems. Power systems, at the same time, heavily rely on information technologies to achieve efficient and reliable operation. Recent initiatives to upgrade power systems into smart grids target an even tighter integration with information technologies to enable the integration of renewable energy sources, local and bulk generation and demand response. Thus for the reliable operation of smart grids, it is essential that its information infrastructure is secure and reliable in the face of both failures and attacks. This thesis contributes to improving the security of power systems against attacks on their information infrastructures. The contributions lie in three areas: data integrity, data condentiality, and data availability of power system applications.
We analyze how characteristics of power system applications can be leveraged for detection and mitigation of data integrity attacks. We consider singleand multi-area power system state estimation. For single-area state estimation, we look at the integrity of measurement data delivered over a wide area communication network. We deffine security metrics that quantify the importance of particular components of the communication network, and that allow us to optimize the deployment of network, transport and application layer security solutions. For multi-area state estimation, we look at the integrity of data exchanged between the control centers of neighboring areas in face of a targeted trojan that compromises an endpoint of the secure communication tunnel. We deffine multiple attack strategies and show that they can signifficantly disturb the state estimation. Moreover, we propose schemes that could be used for detection, localization, and mitigation of data integrity attacks.
We investigate how to provide data confidentiality for power system applications when they utilize cloud computing. We focus on contingency analysis and propose an approach to obfuscate information regarding power flows and the presence of a contingency violation while allowing the operator to analyze contingencies with the needed accuracy in the cloud. Our empirical evaluation shows that the errors introduced into power flows due to the proposed obfuscation are small, and that the RMS errors introduced grow linearly with the magnitude of obfuscation.
We study how to improve data availability in face of gray hole attacks combined with traffic analysis. We consider two cases: SCADA substation to control center communication using DNP3, and inter-control center communication. In the first case, we propose a support vector machine-based traffic analysis algorithm that uses only the information on timing and direction of three consecutive messages, and show that a gray hole attack can be effectively performed even if the traffic is sent through an encrypted tunnel. We discuss possible mitigation schemes, and show that a minor modication of message timing could help mitigate the attack. In the second case, we study how anonymity networks can be used to improve availability at the price of increased communication overhead and delay. We show that surprisingly availability is not always improved with more overhead and delay. Moreover, we show that it is better to overestimate than to underestimate the attacker's capabilities when conguring anonymity networks.
Stockholm: KTH Royal Institute of Technology, 2014. , vi, 48 p.