Vulnerability Assessment of N-1 Reliable Power Systems to False Data Injection Attacks

Zhigang Chu, Jiazi Zhang, Oliver Kosut, Lalitha SankararXiv 2019 · OpenReview

Abstract

This paper studies the vulnerability of large-scale power systems to false data injection (FDI) attacks through their physical consequences. Prior work has shown that an attacker-defender bi-level linear program (ADBLP) can be used to determine the worst-case consequences of FDI attacks aiming to maximize the physical power flow on a target line. Understanding the consequences of these attacks requires consideration of power system operations commonly used in practice, specifically real-time contingency analysis (RTCA) and security constrained economic dispatch (SCED). An ADBLP is formulated with detailed assumptions on attacker's knowledge, and a modified Benders' decomposition algorithm is introduced to solve such an ADBLP. The vulnerability analysis results presented for the synthetic Texas system with 2000 buses show that intelligent FDI attacks can cause post-contingency overflows.

Introduction

Modern power grids are managed by Energy Management Systems (EMS) that rely on real-time data from SCADA sensors. To ensure reliability, operators perform Real-Time Contingency Analysis (RTCA) and Security Constrained Economic Dispatch (SCED). This ensures the system is "$N-1$ reliable"—meaning if any one line or generator fails, the rest of the system can carry the load without overloads.

While previous research showed that FDI attacks could disrupt simple grid models, those studies often ignored the complex $N-1$ security constraints used in actual practice. Furthermore, finding the "worst-case" attack on a massive, real-world grid is a computationally hard problem. This paper bridges the gap by evaluating the vulnerability of large-scale, $N-1$ reliable systems against attackers who understand the grid's defensive logic.

Methods: The Attacker-Defender Framework

The authors model the interaction as an Attacker-Defender Bi-Level Linear Program (ADBLP). In this framework, the upper level represents the attacker, who aims to maximize the physical power flow on a specific target line. To maintain stealth and avoid detection by "bad data" eliminators, the attacker minimizes the total magnitude of the data change using an $l_1$-norm penalty as a proxy for sparsity.

The lower level represents the system's automated response. It solves a SCED problem to minimize costs while satisfying $N-1$ reliability constraints. This creates a complex feedback loop where the attacker's manipulated data influences the system's optimization decisions.

Flowchart illustrating the attacker-defender interaction, where the attacker designs false measurements to manipulate the system's RTCA and SCED processes.

To solve this for a large-scale system, the authors introduce a Modified Benders’ Decomposition (MBD) algorithm. Standard Mixed-Integer Linear Programming (MILP) becomes computationally prohibitive as the number of variables grows. The MBD approach breaks the problem into a "Master Problem" (the attacker's move) and a "Slave Problem" (the system's response), iterating between them to find a lower bound for the worst-case attack.

Experimental Setup

The study utilizes a high-fidelity Synthetic Texas System benchmark, consisting of 2,000 buses, 3,210 branches, and 432 generators. The simulation was performed using OpenPA, an open-source Java-based EMS platform, integrated with Matlab and the CPLEX solver. The system was modeled with a 2% net load loss handling and a 90% warning threshold for line flows to reflect realistic operating conditions.

Key Results and the Cyber-Physical Gap

The most dangerous aspect of these attacks is the creation of a "Cyber-Physical Gap"—a discrepancy between the operator's perceived state (the "cyber" view) and the actual physical state of the grid. By re-distributing loads strategically, the attacker tricks the SCED engine into moving the grid into a physically dangerous state while the operator's monitors show everything is within safe limits.

Scatter plot demonstrating the 'Cyber-Physical Gap': physical power flows frequently exceed 100% limits (violations) while cyber-perceived flows remain within safe bounds.

In the 2,000-bus Texas system, the researchers tested attacks targeting 25 different contingency warnings. They successfully caused physical overflows in 8 out of 25 cases. In one specific instance targeting line ‘ln-2025-2055’, the attacker predicted a flow of 112.2% of the limit. While the physical system experienced this overflow, the operator’s cyber view showed the line at a safe 90-100% capacity.

Conclusion

This research highlights a significant vulnerability in large-scale power systems. Even with $N-1$ reliability measures, sophisticated FDI attacks can induce physical damage. While the attacker requires significant "insider" knowledge (topology, costs, and algorithms), the results provide a necessary upper bound for system risk assessment. Future work will focus on defense strategies, such as identifying and encrypting measurements on the most vulnerable branches identified by this assessment tool.

Summary updated

Abstract

This paper studies the vulnerability of large-scale power systems to false data injection (FDI) attacks through their physical consequences. Prior work has shown that an attacker-defender bi-level linear program (ADBLP) can be used to determine the worst-case consequences of FDI attacks aiming to maximize the physical power flow on a target line. Understanding the consequences of these attacks requires consideration of power system operations commonly used in practice, specifically real-time contingency analysis (RTCA) and security constrained economic dispatch (SCED). An ADBLP is formulated with detailed assumptions on attacker's knowledge, and a modified Benders' decomposition algorithm is introduced to solve such an ADBLP. The vulnerability analysis results presented for the synthetic Texas system with 2000 buses show that intelligent FDI attacks can cause post-contingency overflows.
BibTeX
@misc{chu2019vulnerability,
  author = {Zhigang Chu and Jiazi Zhang and Oliver Kosut and Lalitha Sankar},
  title = {{Vulnerability Assessment of N-1 Reliable Power Systems to False Data Injection Attacks}},
  year = {2019},
  eprint = {1903.07781},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1903.07781},
}
Figure