Adversarially Robust Intrusion Detection for PMU-Enhanced Power Grids: Detecting Stealthy Attacks in a Wide Area Control Framework
Conference
Asif, M, Mali, P, Haider, MZ et al. (2025). Adversarially Robust Intrusion Detection for PMU-Enhanced Power Grids: Detecting Stealthy Attacks in a Wide Area Control Framework
. IEEE POWER AND ENERGY SOCIETY GENERAL MEETING 2010, 10.1109/PESGM52009.2025.11225385
Asif, M, Mali, P, Haider, MZ et al. (2025). Adversarially Robust Intrusion Detection for PMU-Enhanced Power Grids: Detecting Stealthy Attacks in a Wide Area Control Framework
. IEEE POWER AND ENERGY SOCIETY GENERAL MEETING 2010, 10.1109/PESGM52009.2025.11225385
Phasor measurement units (PMUs) have enhanced WAMPAC-based grid operations but also exposed them to optimized false data injection attacks (OFDIAs) that manipulate PMU data to disrupt stability and evade conventional bad data detection (BDD). Our research shows that machine learning-based intrusion detection systems (IDSs) - including RNN-LSTM and CNN-LSTM - are vulnerable not only to common adversarial methods (FGSM, BIM, PGD, MIM) but also to PMU-specific attacks exploiting temporal and spatial data dependencies. To address these challenges, we propose a novel IDS that integrates generative adversarial networks (GANs) with long short-term memory (LSTM) models within the WAMPAC framework. We formalize PMU-specific adversarial attacks, demonstrate the vulnerabilities of existing IDS models under these scenarios, and develop a hybrid GAN-LSTM architecture capable of detecting both optimized and adversarial attacks. Validation using the IEEE 39-bus system with real-time simulation on OPAL-RT shows that our GAN-LSTM IDS effectively detects adversarial and optimized OFDIAs, significantly improving detection accuracy over state-of-the-art models. Our results highlight the critical importance of adversarial robustness in securing PMU-enhanced power grids against evolving threats.
