Robust Graph Autoencoder-Based Detection of False Data Injection Attacks Against Data Poisoning in Smart Grids
Robust Graph Autoencoder-Based Detection of False Data Injection Attacks Against Data Poisoning in Smart Grids
复制标题
DOI：
10.1109/tai.2023.3286831
复制
发表时间：
2024-03
期刊：
影响因子：
--
通讯作者：
Abdulrahman Takiddin;Muhammad Ismail;R. Atat;K. Davis;E. Serpedin
中科院分区：
文献类型：
--
作者：
Abdulrahman Takiddin;Muhammad Ismail;R. Atat;K. Davis;E. Serpedin
智能电网中基于机器学习的虚假数据注入攻击（FDIA）检测依赖于标记的测量数据进行训练和测试。大多数现有的检测器都是假设训练所采用的数据集具有正确的标记信息而开发的。然而，这样的假设并不总是有效的，因为训练数据可能包括被错误地标记为良性的测量样本，即以前未检测到的对抗性数据中毒样本。忽略这一点会使检测器容易受到数据中毒的影响。我们的调查显示，现有的检测器的检测率（DR）显着恶化高达$\text{9}\text{--}\text{29}{\%}$时，受到数据中毒在广义和拓扑特定的设置。因此，我们提出了一个广义的基于图神经网络的异常检测器，对FDIA和数据中毒是强大的。它只需要良性的数据集进行训练，并采用具有注意力机制的Chebyshev图卷积递归层的自动编码器来捕获测量数据中的空间和时间相关性。所提出的卷积递归图自动编码器模型在各种拓扑结构（来自14，39和118总线系统）上进行了训练和测试。由于这些因素，它产生了稳定的广义检测性能，在DR中针对未观察到的拓扑中的高级别数据中毒和不可见FDIA仅降低$\text{1.6}\text{--}\text{3.7}{\%}$。
Machine learning-based detection of false data injection attacks (FDIAs) in smart grids relies on labeled measurement data for training and testing. The majority of existing detectors are developed assuming that the adopted datasets for training have correct labeling information. However, such an assumption is not always valid as training data might include measurement samples that are incorrectly labeled as benign, namely, adversarial data poisoning samples, which have not been detected before. Neglecting such an aspect makes detectors susceptible to data poisoning. Our investigations revealed that detection rates (DRs) of existing detectors significantly deteriorate by up to $\text{9}\text{--}\text{29}{\%}$ when subject to data poisoning in generalized and topology-specific settings. Thus, we propose a generalized graph neural network-based anomaly detector that is robust against FDIAs and data poisoning. It requires only benign datasets for training and employs an autoencoder with Chebyshev graph convolutional recurrent layers with attention mechanism to capture the spatial and temporal correlations within measurement data. The proposed convolutional recurrent graph autoencoder model is trained and tested on various topologies (from 14, 39, and 118-bus systems). Due to such factors, it yields stable generalized detection performance that is degraded by only $\text{1.6}\text{--}\text{3.7}{\%}$ in DR against high levels of data poisoning and unseen FDIAs in unobserved topologies.