Smart Grid Security and Privacy: From Conventional to Machine Learning Issues (Threats and Countermeasures)

Abstract

Smart Grid (SG) is the revolutionised power network characterised by a bidirectional flow of energy and information between customers and suppliers. The integration of power networks with information and communication technologies enables pervasive control, automation and connectivity from the energy generation power plants to the consumption level. However, the development of wireless communications, the increased level of autonomy, and the growing sofwarisation and virtualisation trends have expanded the attack susceptibility and threat surface of SGs. Besides, with the real-time information flow, and online energy consumption controlling systems, customers’ privacy and preserving their confidential data in SG is critical to be addressed. In order to prevent potential attacks and vulnerabilities in evolving power networks, the need for additional studying security and privacy mechanisms is reinforced. In addition, recently, there has been an ever-increasing use of machine intelligence and Machine Learning (ML) algorithms in different components of SG. ML models are currently the mainstream for attack detection and threat analysis. However, despite these algorithms’ high accuracy and reliability, ML systems are also vulnerable to a group of malicious activities called adversarial ML (AML) attacks. Throughout this paper, we survey and discuss new findings and developments in existing security issues and privacy breaches associated with the SG and the introduction of novel threats embedded within power systems due to the development of ML-based applications. Our survey builds multiple taxonomies and tables to express the relationships of various variables in the field. Our final section identifies the implications of emerging technologies, future communication systems, and advanced industries on the security and privacy issues of SG.

Keywords

• Smart grid (SG)
• security
• privacy
• threats
• machine learning (ML)
• adversarial machine learning (AML)