GRID-MRI: Resilience Assessment of Power Systems under Coordinated Cyber-Physical Attacks

arXiv:2611.17094 · eess.SY, cs.CR, cs.SY · Submitted 2026-11-12 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: We're looking at the paper: "GRID-MRI: Resilience Assessment of Power Systems under Coordinated Cyber-Physical Attacks".

Dev: Here's the gist of the paper. The paper proposes the Grid Multidimensional Resilience Index (GRID-MRI) to assess power system resilience by integrating endogenous disruption severity, non-additive coordination effects identified through counterfactual simulations, and exogenous climatic and economic-regulatory conditions The application of GRID-MRI to a modified IEEE 39-bus system under three escalating cyberphysical attack scenarios reveals that coordinated attacks induce system collapse despite stable individual and additive responses.

Rosa: Framework Development The GRID-MRI framework integrates five resilience dimensions: i) physical, ii) cyber, and iii) operational representing endogenous disruptions and iv) climatic and v) economic-regulatory factors representing exogenous factors.

Dev: The index distinguishes aggregate disruption severity, non-additive coordination effects identified through counterfactual simulations, and scenario-specific climatic and economic-regulatory amplification.

Rosa: Endogenous Severity Quantification The endogenous severity term D¯i gives equal weight to operational response and to combined physical and cyber exposure, defined as D¯i = 1/4 (Dphy,i + Dcyb,i) + 1/2Dop,i.

Title and authors: Rosa: So, to recap, this paper proposes the GRID-MRI index as a comprehensive tool for assessing power system resilience by combining internal disruption severity with external factors and coordination effects. It’s designed to help us see what happens when disruptions are not just additive.

Dev: Right. The core idea is that current methods often look at physical damage, cyber damage, operational response, and external conditions like climate or regulation in isolation, which misses how they interact in a real crisis. This paper tries to fix that by building this multidimensional index.

Taro: I’m curious about the specifics of how they define what counts as an "endogenous disruption severity" term, because that seems like the foundation for everything else they build on top of it.

Rosa: They give equal weight to operational response and to a combination of physical damage and cyber damage in this term. It’s defined as one quarter of that combined physical and cyber exposure plus half the operational severity component.

Dev: That means they are treating the immediate system response, like how fast things start reacting, with a bit more emphasis than just the raw physical or cyber damage alone.

Taro: If we look at that, it sounds like they’re trying to capture the idea that the system's ability to cope is as important as the initial damage itself.

Rosa: That’s right. And then they build on that severity measure by adding a term for those non-additive coordination effects, which is where their counterfactual simulations come in handy.

The paper's summary: Dev: So, the paper summarizes that to find those coordination effects, they simulate every single asset being compromised individually and then delay its rotor speed deviation by the time of the attack for each asset.

Rosa: That specific simulation is what they use to calculate an interaction factor. This factor tells us if coordination amplifies or dampens resilience loss compared to just adding up the individual losses linearly.

Taro: So, if that interaction factor is positive, it means coordinated attacks make things much worse than you’d expect from simply adding up the separate failures.

Dev: That's essentially quantifying how much worse a coordinated attack is than a simple additive response model suggests. It’s about moving beyond just summing up the damage to understanding how the attacks work together.

Rosa: And then they scale this whole thing by these external factors—the climatic and economic-regulatory conditions—using scenario-specific scaling factors. These act as multipliers on the overall severity measure.

Taro: So, it’s not just about the internal system dynamics; it’s about how a bad storm or a change in energy policy can make those coordination effects even more dangerous.

The paper's improvements: Dev: One of the key suggested improvements they highlight is this explicit incorporation of scenario-dependent scaling factors for climatic and economic-regulatory dimensions. They treat them as multiplicative factors on the endogenous severity measure.

Rosa: That lets an AI system adjust its resilience metric based on specific conditions, like a predicted winter demand increase or regulatory vulnerabilities in a particular scenario. It makes the assessment dynamic rather than static.

Taro: I think another improvement they point to is using that interaction factor, eta i, to directly quantify the additional degradation caused specifically by the coordination of the attacks themselves.

Dev: That gives us a quantitative measure of how much worse coordinated attacks are compared to a linear superposition of individual attack responses. It moves us away from just saying "it's worse" to saying "it's worse by this amount due to coordination."

Rosa: They also show that the initial increase in resilience loss is heavily driven by the endogenous disruption severity itself, about eighty-one percent. But after that, the subsequent increase is mainly tied to those climatic and economic-regulatory amplifications, about seventy-eight percent.

Conclusion: Dev: So, to wrap up on "GRID-MRI: Resilience Assessment of Power Systems under Coordinated Cyber-Physical Attacks," the authors show that this index provides a quantitative basis for distinguishing the different mechanisms causing resilience degradation in power systems.

Rosa: The main implication is that additive assessments fall short when dealing with coordinated disturbances. This metric helps us identify dominant vulnerabilities when multiple things happen at once.

Taro: It really shows that just looking at the individual components doesn't tell you the whole story; you have to model the interaction between them to get a real picture of system collapse risk.

Dev: And they’ve shown how this index helps us understand exactly what drives that degradation, whether it's the immediate internal severity or those external stressors amplifying things on top of it.

Rosa: It’s a solid framework for assessing these complex failures, and it sets a new standard for how we should think about resilience in interconnected systems.

Taro: Good stuff. Next time, maybe we look at how this framework applies to the real-world testing scenarios they used on that modified IEEE thirty-nine-bus system <ref:2611.17094#pg1,modified IEEE 39-bus system>.

Isaac Ortega Romero, Ioannis Zografopoulos

University of Massachusetts Boston

eess.SY, cs.CR, cs.SY

Submitted: 2026-11-12

Updated: 2026-11-12

Comments: IEEE PESGM 2027, 5 pages

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 82/100

The gist: The paper proposes the Grid Multidimensional Resilience Index (GRID-MRI) to assess power system resilience by integrating endogenous disruption severity…

Key concepts

GRID-MRI
A new metric designed to measure power system resilience by integrating five dimensions: physical, cyber, operational, climatic, and economic-regulatory. It helps distinguish between simple additive damage and complex coordination effects during disruptions.
Endogenous Severity (D¯i)
This term quantifies the internal severity of a disruption by weighting operational response heavily and combining physical exposure with cyber exposure. It captures the core damage caused by the disruption itself.
Non-Additive Coordination Effects (ηi)
This factor measures how coordinated attacks interact. It compares actual resilience loss against a simple sum of individual attack losses, revealing if coordination actually amplifies the system's vulnerability beyond what is expected linearly.
Exogenous Amplification
Climatic and economic-regulatory factors are used as multiplicative scaling factors on the internal severity measure. This adjusts the resilience score based on external stresses like severe weather or regulatory weaknesses in a specific scenario.

Terminology

Summary

The paper proposes the Grid Multidimensional Resilience Index (GRID-MRI) to assess power system resilience by integrating endogenous disruption severity, non-additive coordination effects identified through counterfactual simulations, and exogenous climatic and economic-regulatory conditions The application of GRID-MRI to a modified IEEE 39-bus system under three escalating cyberphysical attack scenarios reveals that coordinated attacks induce system collapse despite stable individual and additive responses. p. 0

Where the paper says it

The main contributions are as follows: • The GRID-MRI formulation that separates endogenous disruption severity, non-additive coordination effects identified through counterfactual simulations, and scenario-dependent climatic and economic-regulatory amplification. • The application of GRID-MRI on a modified IEEE 39- bus system under three escalating cyber-physical attack scenarios informed by the December 2025 Polish incident [12], demonstrating system collapse under coordinated attacks despite stable individual and additive responses.

Framework Development

The GRID-MRI framework integrates five resilience dimensions: i) physical, ii) cyber, and iii) operational representing endogenous disruptions and iv) climatic and v) economic-regulatory factors representing exogenous factors. The index distinguishes aggregate disruption severity, non-additive coordination effects identified through counterfactual simulations, and scenario-specific climatic and economic-regulatory amplification. p. 0

Where the paper says it

The index distinguishes aggregate disruption severity, nonadditive coordination effects, and exogenous amplification. The main contributions are as follows: • The GRID-MRI formulation that separates endogenous disruption severity, non-additive coordination effects identified through counterfactual simulations, and scenario-dependent climatic and economic-regulatory amplification.

Endogenous Severity Quantification

The endogenous severity term D¯i gives equal weight to operational response and to combined physical and cyber exposure, defined as D¯i = 1/4 (Dphy,i + Dcyb,i) + 1/2Dop,i. This formulation combines physical and cyber indices with operational severity to capture the core endogenous disruption measure. p. 2

Where the paper says it

GRID-MRI Formulation The GRID-MRI integrates: i) endogenous severity, ii) non-additive coordination effects, and iii) scenario-specific exogenous conditions. The endogenous severity gives equal weight to operational response and to combined physical and cyber exposure, defined as: D¯ i = 1 4 (Dphy,i + Dcyb,i) + 1 2Dop,i (12) To capture non-additive effects, the coordinated response is compared against a linear superposition reference.

Non-Additive Coordination Effects

To capture non-additive effects, the coordinated response is compared against a linear superposition reference through simulations where each compromised asset b in Bi is simulated in isolation and its rotorspeed deviation is delayed by the attack time of asset b. The interaction factor ηi is defined as the difference between the coordinated and additive resilience losses, where positive values indicate that coordination amplifies the resilience loss relative to the additive reference. p. 2

Where the paper says it

Each compromised asset b ∈ Bi is simulated in isolation, and its rotorspeed (ω) deviation from the baseline is delayed by the attack time of asset b in Si before superposition: ω add g (t) = ω 0 g (t) + X b∈Bi h ω (b) g (t) − ω 0 g (t) i (13) where ω 0 g (t) denotes the baseline ω and ω (b) g (t) the single-asset response aligned with the attack time of asset b in scenario Si . The resilience losses of the coordinated and additive responses, Rloss,i and Radd loss,i, define the interaction factor as: ηi = Rloss,i − Radd loss,i max Rloss,i, Radd loss,i (14) where ηi = 0 if both losses vanish or if the scenario involves a single asset, while positive values indicate that coordination amplifies the resilience loss relative to the additive reference.

Exogenous Amplification

The final GRID-MRIi is adjusted by the interaction factor ηi and scaled by scenario-specific climatic and economic-regulatory dimensions through the formula GRID-MRIi = D¯i(1+ηi)(1+Dclim,i)(1+Decon-reg,i). The climatic and economic-regulatory dimensions act as multiplicative scaling factors on the endogenous severity measure. p. 2

Where the paper says it

The resilience losses of the coordinated and additive responses, Rloss,i and Radd loss,i, define the interaction factor as: ηi = Rloss,i − Radd loss,i max Rloss,i, Radd loss,i (14) where ηi = 0 if both losses vanish or if the scenario involves a single asset, while positive values indicate that coordination amplifies the resilience loss relative to the additive reference. Finally, the severity D¯ i is adjusted by the interaction factor ηi and scaled by the scenario-specific climatic and economicregulatory dimensions: GRID-MRIi = D¯ i(1+ηi)(1+Dclim,i)(1+Decon-reg,i) (15) where the exogenous indices are specified separately from the dynamic simulations and act as multiplicative scaling factors.

Case Study Findings

The case study on the IEEE 39-bus system under three scenarios demonstrates that coordinated attacks produce resilience losses 3.72 and 4.09 times greater than their respective additive estimates in S2 and S3. p. 4

Where the paper says it

ences increase approximately linearly with the compromised PV capacity, coordinated attacks exhibit nonlinear growth, producing resilience losses 3. 09 times greater than their respective additive estimates in S2 and S3.

Furthermore, the GRID-MRI values increase from 0. p. 3
Where the paper says it

The shorter collapse time and greater resilience loss in S3 demonstrate the accelerated degradation associated with the larger coordinated generation loss. Table II summarizes the disruption indices, resilience losses, interaction factors, and resulting GRID-MRI values.

062 to 1.188 and reach 6. p. 0
Where the paper says it

09 times their additive estimates, while GRID-MRI increases from 0. 188 and reaches 6.

081 under additional exogenous stress in the worst-case scenario. p. 2
Where the paper says it

Based on the December 2025 cyberattack on Polish energy infrastructure [12], [23], a state-sponsored adversary targeting both information technology (IT) and OT domains is assumed, and all scenarios share a 25% winter-load increase. Three attack scenarios are considered: i) S1 is a controlinput attack that manipulates the active power reference of a single 190-MW plant, leaving the communication and control of the remaining plants intact, ii) S2 is a multi-vector attack that disrupts communications with six plants, removing operator observability and control forcing disconnection and removing 1,115 MW, and iii) S3 is a worst-case attack scenario where an adversary with unlimited resources disconnects all nine plants in a coordinated sequence, removing 1,540 MW, under additional climatic and economic-regulatory stress.

The logarithmic decomposition shows that the initial increase is predominantly driven by endogenous disruption severity (81%), whereas the subsequent increase is primarily associated with climatic and economic-regulatory amplification (78%). p. 4
Where the paper says it

188 in S2 and reaches 6. The logarithmic decomposition reveals that the initial increase is predominantly driven by endogenous disruption severity (81%), whereas the subsequent increase is primarily associated with climatic and economic-regulatory amplification (78%).

Conclusion

The work establishes a quantitative basis for distinguishing the dominant mechanisms of multidimensional resilience degradation by proposing GRID-MRI. This metric supports the identification of dominant vulnerabilities under coordinated disturbances. p. 4

Where the paper says it

Climatic and economic-regulatory conditions further amplify the assessed resilience impact under the worst-case scenario, accounting for a 78% increase from S2 to S3. By distinguishing the contributions of endogenous severity, coordination effects, and exogenous amplification, the proposed GRID-MRI metric provides a comprehensive characterization of resilience degradation and supports the identification of dominant vulnerabilities under coordinated disturbances.

The findings reveal the limitations of additive assessments in capturing nonlinear degradation induced due to attack coordination. p. 4
Where the paper says it

09 times greater than their respective additive estimates. These findings reveal the limitations of additive assessments in capturing nonlinear degradation induced due to attack coordination.

The proposed index provides a comprehensive characterization of resilience degradation and supports identifying dominant vulnerabilities under coordinated disturbances. p. 4
Where the paper says it

Climatic and economic-regulatory conditions further amplify the assessed resilience impact under the worst-case scenario, accounting for a 78% increase from S2 to S3. By distinguishing the contributions of endogenous severity, coordination effects, and exogenous amplification, the proposed GRID-MRI metric provides a comprehensive characterization of resilience degradation and supports the identification of dominant vulnerabilities under coordinated disturbances.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Add a GRID-MRI metric to power system resilience assessment that integrates physical, cyber, operational, climatic, and economic-regulatory factors to distinguish between additive and non-additive coordination effects in disruptions. This allows AI systems to quantify how coordinated attacks amplify system collapse beyond what individual failures would suggest.

  2. Implement counterfactual simulations within the GRID-MRI framework to explicitly identify non-additive coordination effects by simulating individual asset compromises before superposition. This capability enables the AI to model what-if scenarios to determine if coordinated disruptions induce system collapse despite stable individual and additive responses, as shown in the paper's findings.

  3. Incorporate scenario-dependent scaling factors into the GRID-MRI formulation, where climatic and economic-regulatory dimensions act as multiplicative scaling factors on endogenous severity. This allows AI systems to adjust resilience metrics based on exogenous stressors like winter demand increases or regulatory vulnerabilities in a given scenario.

  4. Develop an AI capability that utilizes the calculated interaction factor eta i to quantify the additional degradation attributable to attack coordination. This provides a quantitative measure of how much worse coordinated attacks are compared to a linear superposition of individual attack responses, moving beyond simple additive assessments.

Abstract

The increasing penetration of inverter-based resources and growing cyber-physical interdependencies expose electrical power systems to disturbances that challenge conventional resilience assessments. Existing approaches often evaluate resilience dimensions independently, overlooking the non-additive effects of coordinated disruptions and the influence of exogenous stressors. To address this gap, this paper proposes a Grid Multidimensional Resilience Index (GRID-MRI) that integrates endogenous disruption severity, coordination effects, and climatic and economic-regulatory conditions. The framework quantifies physical, cyber, and operational degradation, identifies interaction effects through counterfactual (i.e., what-if), simulations of individual and coordinated attacks, and incorporates exogenous stressors as scenario-dependent scaling factors. Its application to a modified IEEE 39-bus system under three escalating cyber-physical attack scenarios, informed by the December 2025 Polish cyberattack incident, reveals that coordinated attacks induce system collapse despite stable individual and additive responses. The corresponding resilience losses are 3.72 and 4.09 times their additive estimates, while GRID-MRI increases from 0.062 to 1.188 and reaches 6.081 under additional exogenous stress. These findings demonstrate the limitations of additive assessments and establish a quantitative basis for distinguishing the dominant mechanisms of multidimensional resilience degradation.

Related papers