Scholarlink

<p>The rapid integration of wireless communication technologies in modern microgrid systems has significantly improved operational flexibility, decentralized control, and real-time monitoring. However, this advancement also introduces critical cybersecurity vulnerabilities, particularly Line Outage Masking (LOM) attacks that can manipulate system topology information and mislead control decisions. To address this challenge, this paper proposes a Lightweight Transformer-Based Edge AI (LTBEA) framework for accurate and low-latency LOM attack detection. The proposed model combines graph-based spatial feature extraction with transformer-based temporal attention mechanisms to effectively capture complex interdependencies among grid nodes and time-series measurements. A lightweight CNN-1D module is employed for local feature extraction, while a GRU-lite unit enhances temporal sequence modeling under dynamic load conditions. The system is deployed on edge computing devices to enable real-time analytics, reduce communication overhead, and ensure scalability in resource-constrained environments. Furthermore, adaptive filtering and normalization techniques improve robustness against noise and measurement uncertainties. The experimental results show that the proposed LTBEA framework has a detection accuracy of 97.8% with a false positive rate of only 1.9% and an average detection latency of 12ms under the noisy wireless communication environment, while having excellent robustness. It is shown that the framework is consistently stronger than traditional detection methods, regardless of the size of the microgrid and the type of attack, and that the Low Rate of False Alarm (LRFA) Line Outage Masking (LOM) attack can be reliably detected with minimal computation. The results validate the feasibility of the LTBEA as an effective and usable cybersecurity solution to boost the operational security, reliability, and resilience of next-generation wireless microgrid networks when they are deployed at the edges of the network.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An edge-AI framework with graph transformer learning for resilient microgrid topology attack identification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月10日

查看全文
  • D. Gowtham Chakravarthy,
  • R. Gopi,
  • K. B. Bhaskar,
  • G. Vennira Selvi

摘要

The rapid integration of wireless communication technologies in modern microgrid systems has significantly improved operational flexibility, decentralized control, and real-time monitoring. However, this advancement also introduces critical cybersecurity vulnerabilities, particularly Line Outage Masking (LOM) attacks that can manipulate system topology information and mislead control decisions. To address this challenge, this paper proposes a Lightweight Transformer-Based Edge AI (LTBEA) framework for accurate and low-latency LOM attack detection. The proposed model combines graph-based spatial feature extraction with transformer-based temporal attention mechanisms to effectively capture complex interdependencies among grid nodes and time-series measurements. A lightweight CNN-1D module is employed for local feature extraction, while a GRU-lite unit enhances temporal sequence modeling under dynamic load conditions. The system is deployed on edge computing devices to enable real-time analytics, reduce communication overhead, and ensure scalability in resource-constrained environments. Furthermore, adaptive filtering and normalization techniques improve robustness against noise and measurement uncertainties. The experimental results show that the proposed LTBEA framework has a detection accuracy of 97.8% with a false positive rate of only 1.9% and an average detection latency of 12ms under the noisy wireless communication environment, while having excellent robustness. It is shown that the framework is consistently stronger than traditional detection methods, regardless of the size of the microgrid and the type of attack, and that the Low Rate of False Alarm (LRFA) Line Outage Masking (LOM) attack can be reliably detected with minimal computation. The results validate the feasibility of the LTBEA as an effective and usable cybersecurity solution to boost the operational security, reliability, and resilience of next-generation wireless microgrid networks when they are deployed at the edges of the network.