[Paper Review] Power Grid Defense Against Malicious Cascading Failure

Paulo Shakarian, Hansheng Lei+1 more|arXiv (Cornell University)|Jan 6, 2014

Smart Grid Security and Resilience24 references16 citations

TL;DR

This paper introduces a game-theoretic model for defending power grids against malicious cascading failures, where an attacker targets substations to maximize outage impact and a defender hardens key nodes to minimize damage. The study formalizes deterministic and mixed strategies, proves NP-hardness in most cases, and shows empirically that attackers benefit more from additional resources than defenders, with minimax defense outperforming simpler load-based strategies under optimal attack conditions.

ABSTRACT

An adversary looking to disrupt a power grid may look to target certain substations and sources of power generation to initiate a cascading failure that maximizes the number of customers without electricity. This is particularly an important concern when the enemy has the capability to launch cyber-attacks as practical concerns (i.e. avoiding disruption of service, presence of legacy systems, etc.) may hinder security. Hence, a defender can harden the security posture at certain power stations but may lack the time and resources to do this for the entire power grid. We model a power grid as a graph and introduce the cascading failure game in which both the defender and attacker choose a subset of power stations such as to minimize (maximize) the number of consumers having access to producers of power. We formalize problems for identifying both mixed and deterministic strategies for both players, prove complexity results under a variety of different scenarios, identify tractable cases, and develop algorithms for these problems. We also perform an experimental evaluation of the model and game on a real-world power grid network. Empirically, we noted that the game favors the attacker as he benefits more from increased resources than the defender. Further, the minimax defense produces roughly the same expected payoff as an easy-to-compute deterministic load based (DLB) defense when played against a minimax attack strategy. However, DLB performs more poorly than minimax defense when faced with the attacker's best response to DLB. This is likely due to the presence of low-load yet high-payoff nodes, which we also found in our empirical analysis.

Motivation & Objective

• To model power grid defense against cascading failures as a two-player game between an attacker and defender.
• To formalize deterministic and mixed strategies for both players under resource constraints.
• To analyze the computational complexity of finding optimal strategies in various scenarios.
• To develop scalable algorithms, including a double-oracle approach for mixed strategies.
• To evaluate the model empirically on a real-world U.S. power grid network and compare defense strategies.

Proposed method

• Model the power grid as an undirected graph with source (generation) and load (consumption) nodes.
• Define edge load using shortest-path betweenness to estimate line stress, based on [8] but extended to include source-load dynamics.
• Introduce a failure operator that removes overloaded edges iteratively, simulating cascading failure.
• Formulate the cascading failure game where attacker maximizes disconnected loads and defender minimizes them.
• Use a capacity margin α to set line capacity as (1+α) × initial load, simulating excess capacity.
• Implement a double-oracle algorithm with greedy heuristics to compute minimax mixed strategies for the defender.

Experimental results

Research questions

• RQ1 What is the computational complexity of finding optimal deterministic strategies for the attacker and defender under different resource constraints?
• RQ2 How do mixed strategies compare to deterministic strategies in terms of payoff and robustness?
• RQ3 Can tractable algorithms be developed for large-scale power grid networks?
• RQ4 How does the defender’s payoff scale with increasing resources compared to the attacker’s?
• RQ5 What role do low-load, high-payoff nodes play in the effectiveness of different defense strategies?

Key findings

• The attacker benefits more from increased resources than the defender; when resources increased from 1 to 6, the attacker’s payoff tripled, while the defender’s improved by only 70%.
• Minimax defense performs comparably to a simple load-based (DLB) defense when facing a minimax attacker, but significantly outperforms DLB when facing the attacker’s best response to DLB.
• The presence of low-load yet high-payoff nodes explains why DLB performs poorly against optimal attacker strategies.
• The double-oracle algorithm’s runtime increases per iteration due to growing linear programs and strategy sets, limiting scalability for large networks.
• Runtime scales linearly with strategy size (R² = 0.90 ± 0.2), but even the longest experiment took 12 days, indicating a need for optimization.
• The model favors the attacker, as he gains more from additional resources than the defender can mitigate with equivalent investment.