Detection of False Data Injection Attacks on Smart Grids Based on A-BiTG Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. There are also some formatting problems with the article, such as the full capitalization of some acronyms in the abstract and body, and some not capitalized. Keywords are not followed by a space after the semicolon. There appears to be an arrow missing from the right-hand flowchart in Figure 2.

2.  In the experimental simulation and results analysis section, the variance of the measurement noise is too large. The standard error of the measurement system in a real power system is about 0.5% to 2% of the normal measurement range. Therefore, the authors need to clarify the feasibility of the measurement data.

3. The article should detail how the hyperparameters of all models are tuned, such as the number of layers and neurons, in order to make the different comparison algorithms fair in terms of comparison experiments.

4. In the present, the construction of FDIA needs to satisfy several constraints, such as not to directly tamper with the measured value of the generator; the zero-injection node of the power system, which remains a zero-injection node after the attack; and if the loads are modified, the modified load measured value should be in the range of 50%~150% of the actual value. Therefore the authors need to specify the state of the attack and the corresponding measured values. In addition, the authors need to specify the magnitude of the change in the attacked state and whether it is a single-point attack or a multi-point attack.

 

5. The paper only compares the detection accuracy of different methods and does not elaborate on the detection time of the proposed algorithm, so the authors are advised to compare the detection time of the proposed algorithm with other methods.

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Please refer to the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The proposed paper looks like high quality paper in which the A-BiTG model and assesses its effectiveness across various attack intensities and noise environments is introduced. The paper additionally investigates the limitations of the current approaches and suggests possible areas for future research, such as attack localization and defense mechanisms.

Maybe this question would be more suitable for the editor, but I think that MDPI Electronics might not be the most suitable journal for this paper. This paper is missing some key elements that are usually found in such publications, such as the design, analysis, or application of electronic circuits, devices, or systems. In light of this, I believe that MDPI Energies might be a more suitable option. Perhaps the authors could provide an explanation for their choice of MDPI Electronics journal.

The paper presents a potential solution to a gap in the field by introducing the A-BiTG model. This model combines BiTCN and BiGRU to extract temporal features for detecting FDI attacks in smart grids.

Explanation of Evaluation Metrics: Provide a concise rationale for selecting the evaluation metrics (Accuracy, Precision, Recall, F1-Score) in the methodology section. Having a clear understanding of the advantages and limitations of each metric will enable readers to interpret the results more effectively.

More information regarding the particular attack types—zero-mean, random walk, etc.—used to assess the model under FDI attacks would be beneficial. This makes it possible to evaluate more precisely and lets next researchers compare the A-BiTG model's performance across various attack scenarios.

Reference list looks appropriate to the paper and most of it newer than 5 years old.

Author Response

Please refer to the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The introduction is very long; splitting the first section into an introduction and a related works section would make the text more readable. The description of the main contribution can be inserted into the introduction section, focusing on the aim of the research, rather than the details of the proposed solution.

The paper proposes a solution for an extensively investigated topic. To add some value to the research, the authors could at least introduce the attack model, or provide a literature review of real use cases, or discuss the feasibility of the simulated attack in a real use-case scenario.

The authors should also include some novel approaches in the state of the art. Autoencoder-based algorithms for anomaly detection are a widely used solution, take for example DOI 10.3390/s22103933. Also, physics-informed neural network is a promising field of research, take for example DOI 10.1109/ACCESS.2023.3347989

Author Response

Please refer to the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors did a good job of addressing the questions I asked and recommend accepting the paper in its current version.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors improved the quality of the paper. Still, the authors did not answer to my comment 2, which was:
"The paper proposes a solution for an extensively investigated topic. To add some value to the research, the authors could at least introduce the attack model, or provide a literature review of real use cases, or discuss the feasibility of the simulated attack in a real use-case scenario."
The problem here is not that generally attacks on electrical grids are feasible and real (we know). The point is that the problem of state estimation is connected to the attack model. I know that the paper proposes an algorithm for detecting the attack regardless of network architectures, protocols, attacks and so on, but still, that element could be not completely ignored, because in a real scenario, attacks are not always feasible, so the risk depends on the scenario.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf