Advances, Systems and Applications
Table 2 Comparison of IDS Techniques for SGs
From: A fog-edge-enabled intrusion detection system for smart grids
Ref. | Technique | Advantages | Disadvantages |
|---|---|---|---|
[72] | Transformer and feature extraction layers | High accuracy, low false positives | Centralized architecture, data dependency |
[73] | Autoencoder-GAN architecture | Low false positive rate, high accuracy, high true positive rate | Centralized architecture, model complexity |
[74] | DNN | Real-time detection, feature learning | High-dimensional input, overfitting |
[75] | FL with gradient privacy-preserving quantization | Higher detection accuracy, privacy-preserving | Communication efficiency, convergence |
[76] | Fine-tuned DNN in FL | Improved sensitivity, continuous monitoring | Centralized architecture, scalability |
[77] | CLAIRE with CNNs | Improved accuracy and performance | Complex model, training data requirements |
[78] | FL-based IDS | Protects sensitive data, collaborative training | Lack of authentication and encryption, convergence |
[79] | FL-based IDS | No central data repository, behavior pattern identification | Privacy-preserving, communication overhead |