A Novel Deep Learning-Based Intrusion Detection System for IoT Networks
Round 1
Reviewer 1 Report
1- draw a flowchart explain the complete model
2- give the full details of the dataset used to evalute the suggested model
3- evaluation metrics such as detection rate are very important to show the seternth of the suggested model
4-enrich the paper by citing most recent papers in IDS such as
https://mesopotamian.press/journals/index.php/CyberSecurity/article/view/8
https://mesopotamian.press/journals/index.php/CyberSecurity/article/view/1
https://mesopotamian.press/journals/index.php/CyberSecurity/article/view/7
5- moreover there are many grammatical and formatting errors
Author Response
Point 1: draw a flowchart explain the complete model
Response to Point 1: I would like to express my gratitude to the respected reviewer for this insightful comment. I have prepared an overview diagram of the entire model with directional arrows. The diagram gives an idea about the proposed methodology. And the directional arrows show the flow of the operations associated with the proposed approach.
Point 2: give the full details of the dataset used to evaluate the suggested model
Response to Point 2: I would like to especially thank the respected reviewer for this review. It helped me a lot to improve the quality of my paper. I agree that without the dataset description, the readers may not understand the paper fully. I have added a new paragraph describing the dataset.
Point 3: evaluation metrics such as detection rate are very important to show the seternth of the suggested model
Response to Point 3: I agree with the respected reviewer. The Intrusion Detection Rate (IDR) is an important evaluation metric of the proposed system, which has been overlooked in the submitted paper. However, I have developed and presented the mathematical model in equation 16 to calculate the IDR and presented the IDR of the proposed system in table 2 and the new changes have been reflected in the manuscript.
Point 4: enrich the paper by citing most recent papers in IDS such as
https://mesopotamian.press/journals/index.php/CyberSecurity/article/view/8
https://mesopotamian.press/journals/index.php/CyberSecurity/article/view/1
https://mesopotamian.press/journals/index.php/CyberSecurity/article/view/7
Response to Point 4:
I would like to appreciate the respected reviewer for such a valuable suggestion. I have studied the suggested papers and analyzed the relevancy of the proposed methodology. It has been discovered that the suggested papers are very resourceful and the ideas align with the proposed paper. I have cited these three papers [22, 30] in a different part of the paper which has been marked on the modified paper.
Point 5: moreover there are many grammatical and formatting errors
Response to Point 5: I have carefully reviewed the entire paper and corrected the grammatical mistakes and formatting errors. The modification has been highlighted in the revised paper. I would like to thank the respected reviewer for this suggestion.
Reviewer 2 Report
- The paper is well-written and well-structured. In addition, the topic is interesting.- The author may include a summary of the experimental results in the abstract.
- She also needs to mention which dataset was used. - At the end of the introduction, the author may include a new figure that graphically illustrates the proposed approach. - Section 2 needs to be summarized in tabular form and emphasize the originality of the proposed methodology. - Some interesting works need to be covered in the related work section like: 1. https://www.sciencedirect.com/science/article/abs/pii/S0045790622000337 2. https://link.springer.com/book/10.1007/978-3-030-89025-4 - Some sections are too short (e.g., Section 3.1) and may be merged together. - Please add a short paragraph that describes the adopted dataset. - Line 162: "using algorithm 3.1.2.1." ==> There is a problem with this reference and with the presentation of this algorithm. - Line 246: "A six-layer deep neural network" ====> Please provide arguments about the use of this number of layers. - Line 246: "optimized for to detect" ====> English mistake - Line 247: "We used a Fully Connected Feed Forward Neural Network" ===> Please argue more about the choice of this type of networks - Please explain how the proposed approach may be adopted in a real context to detect different types of attacks. - The author needs to indicate the limitations of the proposed approach and to identify some future work directions.
Author Response
Point 1: The author may include a summary of the experimental results in the abstract.
Response to Point 2: I would like to thank the respected reviewer for this valuable suggestion. I have added the summary of the experimental results in the abstract. Now the abstract sounds more interesting. It improved the quality of the paper and the credit goes to the keen observation and valuable suggestion of this respected reviewer.
Point 2: She also needs to mention which dataset was used.
Response to Point 2: I would like to thank the respected reviewer for this review especially. It helped me a lot to improve the quality of my paper. I agree that without the dataset description, the readers may not understand the paper thoroughly. I have added a new paragraph describing the dataset and the new changes have been highlighted in the manuscript.
Point 3: At the end of the introduction, the author may include a new figure that graphically illustrates the proposed approach.
Response to Point 3: I would like to express my gratitude to the respected reviewer for this insightful comment. I have prepared an overview diagram of the entire model with directional arrows. The diagram gives an idea about the proposed methodology. And the directional arrows show the flow of the operations associated with the proposed approach.
Point 4: Section 2 needs to be summarized in tabular form and emphasize the originality of the proposed methodology.
Response to Point 4: Thank you for this valuable comment. I have created a list in the introduction of the paper where the core and unique contribution of the paper has been highlighted. These contributions refer to the originality of the paper.
Point 5: Some interesting works need to be covered in the related work section like:
- https://www.sciencedirect.com/science/article/abs/pii/S0045790622000337
- https://link.springer.com/book/10.1007/978-3-030-89025-4
Response to Point 5: I would like to appreciate this valuable suggestion and I totally agree with this comment. I have found insightful direction on Distributed Denial of Service Attacks which has been addressed in the performance analysis section with proper citation (section 4.3.2).
The second suggested paper helped me to understand the state-of-the-art deep learning applications in IoT security and privacy. It has been discussed and cited in the last paragraph of the literature review.
Point 6: Some sections are too short (e.g., Section 3.1) and may be merged together.
Response to Point 6: Section 3.1 has been elaboration, and other smaller paragraphs has been extended. I would like to express my gratitude for this suggestion. After brining this suggestion modification, the overall look of the paper has become much better than before.
Point 7: Please add a short paragraph that describes the adopted dataset.
Response to Point 7: I would like to appreciate this comment. I have realized that the paper loses its strength without describing the dataset. I have added a separate section to the paper and described the dataset in that section per the reviewer suggestion.
Point 8: Line 162: "using algorithm 3.1.2.1." ==> There is a problem with this reference and with the presentation of this algorithm.
Response to Point 8: I would like to thank the respected reviewer for this keen observation. I have noticed that the proposed feature extraction algorithm went missing from the submitted paper because of a package error in the LaTeX. I have fixed the error, and now the algorithm is back again in the new version.
Point 9: Line 246: "A six-layer deep neural network" ====> Please provide arguments about the use of this number of layers.
Response to Point 9: Thank you, respected reviewer, for this comment. I have explained the reasons behind using the six-layer deep neural network. After this modification, the quality of the paper has improved and the readers can understand the reasons behind using the proposed network architecture.
Point 10: Line 246: "optimized for to detect" ====> English mistake
Response to Point 10 This is a very keen observation. I would like to specially thank the reviewer for this comment. I have corrected the grammatical mistake in the revised paper.
Point 11: Line 247: "We used a Fully Connected Feed Forward Neural Network" ===> Please argue more about the choice of this type of network
Response to Point 11: This is an excellent observation. I would like to express my gratitude to the reviewer for addressing this issue. I have explained the reasons behind using a Fully Connected Feed Forward Neural Network architecture. This additional explanation has given the paper a better shape. Thank you, respected reviewer, for your valuable comment.
Point 12: Please explain how the proposed approach may be adopted in a real context to detect different types of attacks.
Response to Point 12: I am glad to receive this revision request. I have explained the real-world impact of the five types of intrusion presented in this paper. After a short discussion of the attacks and how they harm in the real world, then the analysis was performed using the confusion matrix. As a result, it gives an idea of how the proposed approach be adopted in the real context.
Point 13: The author needs to indicate the limitations of the proposed approach and to identify some future work directions.
Response to Point 13: This is a very valuable comment. In response to this comment, I have included a new section in the paper titled ‘Limitation & Future Scope.’ After modifying the manuscript according to this comment, the paper quality has improved significantly. The proposed methodology has future direction, and the subsequent research will be continued.
Reviewer 3 Report
Authors have proposed a novel Deep Learning (DL)-based intrusion detection system for IoT devices.
Following comments can be added in the main text as follows:
A. In the Introduction section, authors need to declare main contributions exactly for intrusion detection strategy.
B. Authors can discuss on following case studies as a new relevant topics in Section 2:
- Recurrent deep learning-based feature fusion ensemble meta-classifier approach for intelligent network intrusion detection system.
- Improved intrusion detection method for communication networks using association rule mining and artificial neural networks.
- Improved intrusion detection method for communication networks using association rule mining and artificial neural networks.
- Extreme learning machine and bayesian optimization-driven intelligent framework for IoMT cyber-attack detection.
C. The authors can explain each Confusion Matrix related to existing attacks with details.
D. According to Table 2, which algorithm was applied for evaluation of prediction factors ??
E. Please explain future work in Conclusion section.
Author Response
Point 1: In the Introduction section, authors need to declare main contributions exactly for intrusion detection strategy.
Response to Point 1: I am truly grateful for this valuable comment. I have modified the manuscript according to this comment. The contributions have been listed in the introduction.
Point 2: Authors can discuss on following case studies as new relevant topics in Section 2:
- Recurrent deep learning-based feature fusion ensemble meta-classifier approach for intelligent network intrusion detection system.
- Improved intrusion detection method for communication networks using association rule mining and artificial neural networks.
- Extreme learning machine and bayesian optimization-driven intelligent framework for IoMT cyber-attack detection.
Response to Point 2: These suggestions have become very helpful in improving the quality of the paper. I would like to express my gratitude to the reviewer for making these valuable suggestions. I have discussed the first two suggested paper and compared them with the proposed paper in section two. The third paper has been addressed in the introduction where there is a discussion on the Medical Internet of Things.
Point 3: The authors can explain each Confusion Matrix related to existing attacks with details.
Response to Point 3: I would like to thank the respected reviewer for this valuable comment. I have explained the confusion matrix related to the existing attacks with more details. It has improved the quality of the paper and the performance analysis has become easier to understand.
Point 4: According to Table 2, which algorithm was applied for evaluation of prediction factors ??
Response to Point 4: I used the deep neural network with Gradient Descent algorithm to update weights of the hidden nodes. The values listed on table 2 has been generated from the prediction from the trained network.
Point 5: Please explain future work in Conclusion section.
Response to Point 5: I added the future direction of this research in the conclusion. I am truly grateful to the reviewer for this comment. After adding the future work in the conclusion, the paper has become more interesting than before.
Reviewer 4 Report
Although IDS based on the DNN is proposed in the paper, it has the following limitations to be resolved.
First, sinkhole, wormhole, blackhole attacks have been investigated for long time and thus many related work have been developed to detect these attacks. Authors need to clarify why the DNN-based scheme is needed against the related work.
Second, there are no comparison results to other related work in the paper. It needs to be demonstrated that the proposed scheme outperforms the related work.
Author Response
Although IDS based on the DNN is proposed in the paper, it has the following limitations to be resolved.
Point 1: First, sinkhole, wormhole, blackhole attacks have been investigated for long time and thus many related work have been developed to detect these attacks. Authors need to clarify why the DNN-based scheme is needed against the related work.
Respond to Point 1: I am truly glad to get this valuable review from the respected reviewer. This comment kept me thinking for a couple of days. I did more literature review and discovered that the traditional rule-based method is becoming less effective for the intrusion model proposed in the paper. The massive amount of data and availability of high-speed internet connection has created a more drastic version of these attacks. Deep neural networks are a good fit for big data analysis, discovering patterns, and predicting from massive amounts of data. That is why I have used DNN. A new paragraph has been added to the paper in response to this comment.
Point 2: Second, there are no comparison results to other related work in the paper. It needs to be demonstrated that the proposed scheme outperforms the related work.
Respond to Point 2: I agree with the reviewer that the paper has a weakness, and that is the comparison with other papers. This issue delayed this paper from submitting for 4 weeks. I kept no stone unturned to find papers that use similar evaluation criteria and dataset features. However, it went in vain. The recently published review papers do not have enough resources to present some logical comparison of the proposed paper with papers.
Reviewer 5 Report
Alot of work is needed for this paper to be accepted for publication in my opinion, other than the extensive English editing required, the paper has low significance and novelty. The related work needs to be rewritten to include the similar methods including their weaknesses and strengths. The methology it self need to be more clear, mYbe using a bulletin to highlight the steps. Section 313 is wrong please correct. The verification is accepted but can be improved. Abstract , introduction and conclusion needs to be elaborated and explainmore.
Author Response
Point 1: A lot of work is needed for this paper to be accepted for publication in my opinion, other than the extensive English editing required, the paper has low significance and novelty.
Response to Point 1: I would like to thank the reviewer for this motivational review which encouraged me to revise the paper multiple times. Through this revision, I have corrected the grammatical and syntactical issues. At the same time, I highlighted the significance and novelties in different paragraphs in the revised versions of the paper.
Point 2: The related work needs to be rewritten to include similar methods, including their weaknesses and strengths.
Response to Point 2: I am truly grateful for such a lovely review. I have analyzed the similar methodologies presented in the literature review and highlighted the weaknesses and strengths of those systems.
Point 3: The mythology itself need to be more clear, my be using a bulletin to highlight the steps.
Response to Point 3: I would like to express my gratitude to the reviewer for this keen observation. I agree with the response. I have prepared a graphical illustration of the overall methodology to make it clearer.
Point 4: Section 313 is wrong please correct.
Response to Point 4: I agree with reviewer 5 that the mentioned section creates ambiguity and from different perspectives, it seems incorrect. I have modified this section and explained the working principle of the Mitigation Module with the help of a mathematical model, which eliminates ambiguity.
Point 5: The verification is accepted but can be improved.
Response to Point 5: I would like to thank the reviewer for this comment. It helped me to further improve the quality of this paper. I have added an additional evaluation metric, Intrusion Detection Rate (IDR), to improve the quality of verification method presented in the paper.
Point 6: Abstract, introduction and conclusion needs to be elaborated and explain more.
Response to Point 6: I would like to express a heartiest gratitude for this valuable comment. I have elaborated on the abstract and intrusion. I also attempted to explain more in the conclusion by adding the future scope and limitations. However, it makes the conclusion too lengthy. That is why I have added a new section titled ‘Limitations & Future Scope.’ Because of this comment by the reviewer, the paper’s quality has improved significantly. I am truly grateful for this respected and knowledgeable reviewer’s wise comments. It helped me a lot to improve the quality of the paper.
Round 2
Reviewer 2 Report
No more comments to add. Good luck.
Reviewer 5 Report
I The paper improved significantly, I can not see the Future scope and limitation section, is it after conclusion or before?
Please stop over thanking the reviewers, we just doing our job.
