[Paper Review] Protecting the Grid against IoT Botnets of High-Wattage Devices

Saleh Soltan, Prateek Mittal+1 more|arXiv (Cornell University)|Aug 11, 2018

Smart Grid Security and Resilience29 references4 citations

TL;DR

This paper proposes the SAFE and IMMUNE algorithms to compute robust generator operating points in economic dispatch that prevent line overloads after MAD (Manipulation of Demand) attacks by IoT botnets of high-wattage devices. It introduces $αD$-robustness to quantify resilience and provides efficient polynomial-time methods to verify if line overloads can be cleared during secondary control, ensuring grid stability under attack scenarios.

ABSTRACT

We provide methods to prevent line failures in the power grid caused by a newly revealed MAnipulation of Demand (MAD) attacks via an IoT botnet of high-wattage devices. In particular, we develop two algorithms named Securing Additional margin For generators in Economic dispatch (SAFE) Algorithm and Iteratively MiniMize and boUNd Economic dispatch (IMMUNE) Algorithm for finding robust operating points for generators during the economic dispatch such that no lines are overloaded after automatic primary control response to any MAD attacks. In situations that the operating cost of the grid in a robust state is costly (or no robust operating points exist), we provide efficient methods to verify--in advance--if possible line overloads can be cleared during the secondary control after any MAD attacks. We then define the $αD$-robustness notion for the grids indicating that any line failures can be cleared during the secondary control if an adversary can increase/decrease the demands by $α$ fraction. We demonstrate that practical upper and lower bounds on the maximum $α$ for which the grid is $αD$-robust can be found efficiently in polynomial time. Finally, we evaluate the performance of the developed algorithms and methods on realistic power grid test cases. Our work provides the first methods for protecting the grid against potential line failures caused by MAD attacks.

Motivation & Objective

• Address the growing threat of MAD attacks via IoT botnets of high-wattage devices that can destabilize power grids by abruptly altering demand.
• Develop computationally efficient methods to find robust generator operating points that prevent line overloads after primary frequency control response to any MAD attack.
• Provide tools to verify in advance whether temporary line overloads from MAD attacks can be cleared during secondary control, ensuring system resilience.
• Introduce the concept of $αD$-robustness to quantify the maximum allowable demand perturbation ($\alpha$) for which the grid remains secure under MAD attacks.
• Ensure that the proposed methods are scalable and efficient, with polynomial-time complexity for practical deployment on real-world test systems.

Proposed method

• Develop the SAFE algorithm, which solves a single Linear Program (LP) to compute robust generator operating points that prevent line overloads after any MAD attack.
• Propose the IMMUNE algorithm, an iterative method that converges to lower-cost robust operating points than SAFE by solving a sequence of LPs with adaptive updates.
• Define $αD$-robustness as a formal criterion indicating that all line overloads caused by an adversary increasing/decreasing demand by a fraction $\alpha$ can be cleared during secondary control.
• Introduce $\hat{\alpha}$, a computable upper bound for $\alpha^{\max}$, derived from a single LP, to efficiently estimate the maximum safe $\alpha$.
• Develop $\alpha^{(\beta)}$ and $\alpha^{(\gamma,\beta)}$ as tighter lower bounds on $\alpha^{\max}$, computed via iterative LPs with adaptive step size $\lambda$ to improve convergence and accuracy.
• Use a brute-force search and numerical evaluation on IEEE test systems to validate the efficiency and accuracy of the bounds and algorithms.

Experimental results

Research questions

• RQ1 Can robust generator operating points be computed efficiently in economic dispatch to prevent line overloads after any MAD attack?
• RQ2 How can the grid operator verify in advance whether temporary line overloads from a MAD attack can be cleared during secondary control?
• RQ3 What is the maximum allowable demand perturbation $\alpha$ for which the grid remains secure under MAD attacks, and how can this $\alpha^{\max}$ be efficiently bounded?
• RQ4 Can the proposed methods be applied to real-world power systems with polynomial-time complexity and practical scalability?
• RQ5 How do the proposed bounds ($\hat{\alpha}$, $\alpha^{(\beta)}$, $\alpha^{(\gamma,\beta)}$) compare in accuracy and computational cost across different test systems?

Key findings

• The SAFE algorithm computes robust operating points via a single LP, ensuring no line overloads after any MAD attack, with guaranteed feasibility.
• The IMMUNE algorithm produces lower-cost robust operating points than SAFE by iteratively refining the solution, converging in as few as 3 iterations with optimal step size $\lambda=1.1$.
• The upper bound $\hat{\alpha}$, computed in polynomial time via a single LP, provides a very close estimate of $\alpha^{\max}$ in most test cases, including IEEE 14-bus (f), 14-bus (u), 30-bus, and NE 39-bus systems.
• In the IEEE 57-bus (f) system, $\hat{\alpha} = 0.2$ while $\alpha^{\max} < 0.09$, indicating that $\hat{\alpha}$ is a reliable upper bound even when not tight.
• The lower bounds $\alpha^{(\beta)}$ and $\alpha^{(\gamma,\beta)}$ are significantly more accurate than $\alpha^*$, with $\alpha^{(\gamma,\beta)} = \hat{\alpha}$ in the NE 39-bus system, implying exact recovery of $\alpha^{\max}$.
• For $\alpha = 0.31$ in the IEEE 30-bus system, neither SAFE nor IMMUNE finds a robust operating point, but since $\alpha < \alpha^{\max} = 0.37$, secondary control can clear overloads, confirming the method’s practical utility.