View Project

Norwegian AI Directory

Stability under Adversarial INfluence of Transmission Grids


Securing cyber-physical systems (CPS) is critical not only for ensuring the correct and reliable functioning of both the CPS and its environment, but is also required to prevent physical, bodily, or environmental harm. One critical aspect for any CPS, whether at small scale such as in local Internet of Things or for critical infrastructures such as power networks is the problem of observability. Not only is full observability difficult to achieve and often cost-prohibitive, but measurements or even sensors may be subverted. This can lead to malicious influence on control decisions or denial of controllability as well as indirectly to suboptimal decision support and machine learning by adversarial learning. The problems in obtaining secure state estimates for power systems can hence be considered a particularly challenging case as these are inherently distributed, hard real-time systems that impose stringent requirements on any algorithmic approaches to security and robustness. The proposed research seeks to study a family of state estimation algorithms and their security motivated by new developments in power systems anticipated for the coming decades, namely heavier reliance on phasor measurement units (PMUs) and the increased importance of HVDC interconnections. For the latter, the accuracy and trustworthiness of measurements at such points is disproportionately important to the correct attainment of state estimates and subsequent decisions, and has not been studied thus far, nor has the security of hybrid AC-HVDC models. We therefore wish to study the robustness and stability of constrained state estimators to faults and attacks, emphasising malicious state forcing and convergence-prevention attacks. The primary contributions of the proposed work are the identification of critical measurements and topology information for deliberate attacks against hybrid state estimators together with rapid detection and mitigation mechanisms.

Project leader: Stephen Wolthusen

Started: 2019

Ends: 2023

Category: Universiteter

Sector: UoH-sektor

Budget: 9602000

Institution: Institutt for informasjonssikkerhet og kommunikasjonsteknologi

Address: Gjøvik