Cross-layer impact analysis and a novel security architecture for cyber-physical power system

Feng, Xiaomeng (2025) Cross-layer impact analysis and a novel security architecture for cyber-physical power system. University of Southampton, Doctoral Thesis, 135pp.

Record type: Thesis (Doctoral)

Abstract

With the increasing interdependency of advanced information and communication technologies, power systems are undergoing a rapid transition to cyber-physical power systems (CPPS). This interdependency introduces cross-layer cyber threats that propagate their effects from the cyber layer to the physical layer, disrupting power system operations and potentially causing widespread blackouts. This research investigates two cyber challenges affecting CPPS security from two aspects: (1) degraded communication quality of service (QoS), which compromises data availability, and (2) false data injection attacks (FDIAs), which target data integrity.

Degraded QoS poses a critical cross-layer threat to CPPS by disrupting the timely and accurate transmission of control signals or measurements.
Such disruptions undermine key functions such as frequency, voltage regulation in a cross-layer fashion.
To address this, a novel technique is proposed, comprising (1) a CPPS model for quantitatively analyzing the cross-layer impact of resource allocation on physical states, specifically frequency, voltage, and (2) a multi-objective optimization framework to develop an optimal resource allocation strategy that minimizes disruptions to physical state regulation while enhancing QoS. The proposed strategy achieves a 13.74% reduction in frequency deviation and a 4.57% reduction in voltage deviation in the test system.

Another type of cyberattack, FDIAs, also pose critical cross-layer threats to CPPS by targeting data integrity.
By compromising multiple measurement devices and cooperatively manipulating their measurements, FDIAs can construct stealthy attack vectors that evade residue-based bad data detection (BDD), mislead power system state estimation (PSSE), and ultimately cause market instability and economic losses.
With the increasing integration of electricity markets and carbon trading markets, the cross-layer threats posed by FDIAs are further exacerbated due to additional vulnerabilities in energy price calculation mechanisms.
Traditional approaches that assess economic risks based solely on electricity markets are no longer sufficient.
This research represents the first effort to extend the investigation of economic risks induced by FDIAs beyond the electricity market, incorporating the impacts of carbon emission costs.
Simulations reveal an economic risk increase of up to 201.61 ($/MWh) on a certain transmission line in the PJM test system, compared with the traditional risks assessment only considering electricity costs.

Following the economic risk analysis of FDIA, this research further investigates mitigation strategies by disrupting its stealthiness, which depends on their capability of propagating across the system and manipulating a sufficient number of measurements.
To address this, this research introduces the concept of zero-trust architecture (ZTA) and develops a novel security architecture based on a micro-segmentation technique.
This technique divides measuring devices into finer security segments, restricting lateral attack propagation within the cyber layer while reducing FDIA stealthiness in the physical layer.
To optimize the micro-segmentation strategy, a cyber-physical-BDD-enhancement-metric and a Graph Attention Network (GAT) combined with a reinforcement
learning (RL) algorithm are proposed, evaluating the technique’s effectiveness in enhancing BDD detection capability and mitigating the impact of FDIAs. Simulations demonstrate a significant improvement in the BDD detection rate against FDIAs, increasing from 5.23\% to 94.02\% with the proposed technique.