The Research of Multi-Node Collaborative Compound Jamming Recognition Algorithm Based on Model-Agnostic Meta-Learning and Time-Frequency Analysis
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsPlease also see the attachments.
This work proposed a multi-node collaborative compound jamming recognition algorithm based on time-frequency graph and model-agnostic meta-learning. Using the Omniglot data set, the algorithm proposed by the authors was proved to be more accurate rather than other methods with the limited samples. It is evaluated as a new algorithm for recognizing jamming signals in communication networks. However, for the following reasons, what will be published in the journal is judged to be reviewed in depth by other related experts.
In Figures 1 and 2, it is required to present the validity of a preliminary study on whether all jamming signals could be clearly classified as signals shown in the figures and distinguished. In other words, it is necessary to answer whether it is possible to classify all jamming signals generated in real-world communication networks.
An additional Chapter should be written to explain the excellence and limitations of other algorithms related to the algorithm presented in the Introduction section. In other words, it must be presented as differentiation from existing studies and excellent performance of the proposed algorithm.
The core content of this study is the algorithm in Table 1. There is no particular idea, and it is a simple mapping level. This is why the contribution of the paper is weak. It is not clear whether the algorithm is a heuristic algorithm or a mathematical algorithm for discrimination. If it is just a classification technique by learning, it is not creative compared to other works. The limitations of the research should be identified.
It is required to explain whether the range of parameter values presented in Table 2 is a variable value that may occur in a realistic communication environment.
A description of the parameters presented in Table 3 should be added. A description of whether or not the parameter is a hyper-parameter should also be added. Furthermore, additional review studies are essential to see if other variables should be considered.
It is not clear whether the results in Figure 8 are also applicable to common jamming signals.
Only experimental results for a specific dataset (Omniglot data) exist. To prove the validity of the proposed algorithm, it is essential to prove more realistic experimental data.
It is not advisable to present the results of the Figure at the Conclusion of Chapter 5. The results of the Figure should be presented separately from the experimental results.
There are many explanations that violate the journal’s writing rules. In particular, it should be checked that the format of paper writing for figures and tables is acceptable.
Comments for author File: Comments.pdf
Comments on the Quality of English LanguageEnglish correction is absolutely necessary by referring to the journal;s writing rules.
Author Response
Comments 1:In Figures 1 and 2, it is required to present the validity of a preliminary study on whether all jamming signals could be clearly classified as signals shown in the figures and distinguished. In other words, it is necessary to answer whether it is possible to classify all jamming signals generated in real-world communication networks.
Response 1:We add the description of whether time-frequency image could complete the task of jamming recognition in real communication
Comments 2:An additional Chapter should be written to explain the excellence and limitations of other algorithms related to the algorithm presented in the Introduction section. In other words, it must be presented as differentiation from existing studies and excellent performance of the proposed algorithm.
Response 2:We introduced the advantages and disadvantages of each different algorithms at the Introduction part when we were introducing the existing algorithms. Yet, according to your suggestion, we made a more clear classification and more supplementary information.
Comments 3:The core content of this study is the algorithm in Table 1. There is no particular idea, and it is a simple mapping level. This is why the contribution of the paper is weak. It is not clear whether the algorithm is a heuristic algorithm or a mathematical algorithm for discrimination. If it is just a classification technique by learning, it is not creative compared to other works. The limitations of the research should be identified.
Response 3:In the Introduction part, this article stated some existing the limits of current jamming recognition algorithm. Our main innovations are as follows: 1. Solved the problem of jamming recognition in few-shot condition. 2. Used Omniglot data set to complete the expansion of jamming signal data set. 3. Integrated the idea of multi-nodes collaboration into the composite jamming recognition in few-shot condition, improved composite interference recognition accuracy.
Comments 4:It is required to explain whether the range of parameter values presented in Table 2 is a variable value that may occur in a realistic communication environment.
Response 4:The parameter settings in this paper are consistent with the real world communication environment. We have supplemented related information.
Comments 5:A description of the parameters presented in Table 3 should be added. A description of whether or not the parameter is a hyper-parameter should also be added. Furthermore, additional review studies are essential to see if other variables should be considered.
Response 5:We reintroduced the number of the parameters. These parameters are not hyperparameters. And no other variables needs to be considered after consulting related materials.
Comments 6:It is not clear whether the results in Figure 8 are also applicable to common jamming signals.
Response 6:It does applies to common jamming signals because common jamming signals can also be identified by the line features of time-frequency images.
Comments 7:Only experimental results for a specific dataset (Omniglot data) exist. To prove the validity of the proposed algorithm, it is essential to prove more realistic experimental data.
Response 7:This article used Omniglot data set because it uses the line features of numbers to achieve digital recognition, which is highly similar to jamming recognition through time-frequency graphs. We did not only aimed at Omniglot data set. It just plays a role in expanding the sample.
Comments 8:It is not advisable to present the results of the Figure at the Conclusion of Chapter 5. The results of the Figure should be presented separately from the experimental results.
Response 8:The modification has been completed.
Comments 9:There are many explanations that violate the journal’s writing rules. In particular, it should be checked that the format of paper writing for figures and tables is acceptable.
Response 9:Related modification has been completed.
Reviewer 2 Report
Comments and Suggestions for AuthorsPlease find below some important points that should be improved:
1) Page 1 line 42 "Article [4] proposed an jamming recognition algorithm" should be "Article [4] proposed a jamming recognition algorithm".
2) In the Introduction, authors should better motivate the importance of antijamming algorithms in 5G, beyond 5G and 6G communication networks. In particular, jamming could potentially hinder not only the effectiveness of communications per se, but also of some other crucial services enabled by 5G/6G cellular networks such as localization and synchronization, as well as channel estimation at large. There exists a huge literature mentioning the importance of preventing jamming and similar deleterious attacks in such contexts, particularly in cell-free deployments. By searching on Google Scholar for keywords such as 'Positioning and synchronization in cell-free deployments", a lot of related literature can be found and used to better motivate the discussion in the Introduction.
3) At the end of the Introduction, please consider adding a Table comparing the main pros and cons of the closest existing works and highlighting how the present contribution advances state-of-the-art.
4) In the signal pre-processing stage, authors should better discuss what happens if the samples are correlated due to overlapping windows in the short time Fourier transform (STFT). This is an important issue that has been investigated in the literature (keywords such as 'Time-frequency correlation in spectrogram samples' are used on Google Scholar literature) and needs to be discussed within the framework of the considered problem.
5) Results in Section IV in terms of recognition accuracy could be complemented by a computational complexity analysis aiming at revealing the cost of the proposed algorithm.
Comments on the Quality of English LanguageN/A
Author Response
Comments 1: Page 1 line 42 "Article [4] proposed an jamming recognition algorithm" should be "Article [4] proposed a jamming recognition algorithm".
Response 1:Modification has been completed
Comments 2: In the Introduction, authors should better motivate the importance of antijamming algorithms in 5G, beyond 5G and 6G communication networks. In particular, jamming could potentially hinder not only the effectiveness of communications per se, but also of some other crucial services enabled by 5G/6G cellular networks such as localization and synchronization, as well as channel estimation at large. There exists a huge literature mentioning the importance of preventing jamming and similar deleterious attacks in such contexts, particularly in cell-free deployments. By searching on Google Scholar for keywords such as 'Positioning and synchronization in cell-free deployments", a lot of related literature can be found and used to better motivate the discussion in the Introduction.
Response 2:We add the influence of jamming signals on 6G communication and its additional functions, and discussed it by citing relevant references.
Comments 3: At the end of the Introduction, please consider adding a Table comparing the main pros and cons of the closest existing works and highlighting how the present contribution advances state-of-the-art.
Response 3:We added more clear classification and more detailed information of current algorithms at the Introduction part.
Comments 4: In the signal pre-processing stage, authors should better discuss what happens if the samples are correlated due to overlapping windows in the short time Fourier transform (STFT). This is an important issue that has been investigated in the literature (keywords such as 'Time-frequency correlation in spectrogram samples' are used on Google Scholar literature) and needs to be discussed within the framework of the considered problem.
Response 4:We supplemented the influence of overlapping Windows on the results of time-frequency analysis
Comments 5: Results in Section IV in terms of recognition accuracy could be complemented by a computational complexity analysis aiming at revealing the cost of the proposed algorithm.
Response 5:We supplemented the Params and Flops of the model to reveal the cost of the proposed algorithm.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThere are some limitations to solve the given problem under the realistic environments even though the some modifications are well considered in the revised version.
Truly yours.
Reviewer 2 Report
Comments and Suggestions for AuthorsAuthors applied all the comments.
Comments on the Quality of English LanguageN/A