Novel and Efficient Schemes for Security and Privacy Issues in Smart Grids

The past years have witnessed the fast development of smart grids all over the world. The introduction of digital communication technologies into the power system makes smart grids more efficient and intelligent. In the meantime, however, wide security and privacy concerns arise due to the increasing system complexities. Without proper protection, smart grids is extremely vulnerable to various attacks, such as conventional physical damages and emerging cyber attacks. On the other hand, even tiny system faults, if not detected and resolved in a real time manner, would lead to large scale power outage with unexpected loss. Besides, customer's privacy is also severely threatened by the provision of fine-grained power consumption data in smart grids. Motivated by these concerns, three novel schemes from different technical perspectives are proposed in the dissertation.;For the first topic, an efficient homomorphic encryption-based system was proposed for securing data transmission, data sharing and operations among different parties. In this work, we first proposed a system framework tailored for homomorphic encryption techniques which have great potential to secure data, enable privacy-preserving data sharing and thereby improve the overall efficiency of smart grids. Based on the proposed system framework, we then designed a practical system with an extended partially homomorphic encryption scheme. With homomorphic features, we prove that the designed system well supports privacy-preserving data aggregation and power consumption statistical analysis in smart grids.;For the second topic, a metering scheme was proposed to protect customer's privacy. In this work, a reading distortion scheme was first designed to distort smart meter data in a way that only data senders (i.e. customers) are able to access the original power consumption data. With distorted power consumption data, an aggregated billing mechanism was then proposed to guarantee accurate billing service. For power consumption analysis and prediction, we designed a distribution reconstruction algorithm to recover the original power consumption distribution from distorted power consumption data. To show the security, two potential attacks were investigated theoretically. Experimental results on real world power consumption data were discussed in the end.;For the third topic, a power quality monitoring scheme using change-point detection techniques was investigated. After modeling pre-event and post-event power signal, we proposed a weighted CUSUM algorithm to detect common power quality events, i.e. sags, transients, swells and harmonics. With experimental results, we compared proposed scheme with conventional power quality monitoring techniques and concluded with the superiority of the proposed scheme. We also extend the scheme to distributed version under multi-sensor scenario. The proposed MVW-CUSUM scheme is compared with recent MBQCUSUM scheme in terms of detection latency and robustness.