핵심 개념
The paper proposes a game-theoretic approach to strategically place an additional Phasor Measurement Unit (PMU) in the power grid to enhance the system's resilience against False Data Injection Attacks (FDIAs) by introducing redundancy in attack-susceptible areas.
초록
The paper presents a game-theoretic framework to address the problem of strategically placing an additional PMU in a power grid to improve the system's resilience against False Data Injection Attacks (FDIAs).
The key highlights are:
The authors formulate the interaction between the attacker and the defender as a two-player zero-sum game, where the attacker aims to launch stealthy FDIAs to manipulate the power system state estimation, while the defender's goal is to strategically place an additional PMU to maximize the detection rate of such attacks.
The authors propose a reinforcement learning-based algorithm, namely Exponential Weights for Exploration and Exploitation (EXP3), to compute the Nash Equilibrium (NE) solution of the game without requiring complete knowledge of the opponent's actions and rewards.
The authors evaluate the proposed approach using the IEEE 14-bus system and compare the performance with a naive defense strategy. The results show that the game-theoretic approach increases the FDIA detection rate by 14.85% and 36.69% compared to the naive defense, for systems without and with zero-injection buses, respectively.
The authors demonstrate that the EXP3 algorithm can effectively compute the NE solution, which is comparable to the solution obtained using the Lemke-Howson method that requires full knowledge of the game.
The proposed game-theoretic framework provides a practical and cost-effective solution for power system operators to enhance the resilience of their grids against strategic cyber-attacks by optimally placing additional PMUs.
통계
The paper reports the following key figures:
The number of possible attack actions (|SA|) is P
u∈NP MU (2|Lu|+1 −1), where NP MU is the set of buses without PMUs and Lu is the set of lines connected to bus u.
The probability of detecting a False Data Injection Attack (FDIA) using the proposed game-theoretic approach is 40.75% and 62.50% for the IEEE 14-bus system without and with zero-injection buses, respectively.
The probability of detecting a FDIA using a naive defense strategy (uniform distribution of defense actions) is 25.90% and 23.81% for the IEEE 14-bus system without and with zero-injection buses, respectively.