Machine Learning-Based Intrusion Detection Methods in IoT Systems: A Comprehensive Review

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The review provides full coverage of different machine learning techniques in its discussion, supervised, unsupervised, deep learning, and hybrid models inclusive. This blanket approach will enable readers understand the capability and weakness of different methods used.

By exploring practical applications across diverse domains such as smart homes, healthcare, transportation and industrial automation, the paper illustrates how machine learning based IDS have real-world relevance. 

 

However, scalability for large-scale deployment in IoT systems is a concern when using intrusion detection methods based on machine learning. It is important to guarantee effective resource management and minimal computational overload for successful implementation.

 Improved transparency and trust in the detection process should drive an investigation into interpretable ML approaches.

The paper provides a comprehensive overview of machine learning-based intrusion detection methods in IoT systems, covering various approaches, practical applications, and industry implications. However, addressing minor concerns such as the need for standardized evaluation metrics, scalability considerations, and model interpretability could further enhance the clarity and robustness of the paper.

Author Response

Thank you very much for your comments and constructive suggestions.
In the attached document we have provided a response to each of your comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

1. Please reorganize the introduction by including section 3 and 5.

2. Please compare and evaluate the pros and cons of traditional IDS and ML-based IDS, including section 4.3.

Author Response

Thank you very much for your comments and constructive suggestions.
In the attached document we have provided a response to each of your comments.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have undertaken a comprehensive review of the most recent machine learning-based IDS.

Below are some comments for the authors:

1-Introduction Needs Expansion: The introduction is very limited. Consider expanding it to include:

-Importance of Intrusion Detection Systems (IDS)

-Limitations of traditional IDS methods

-Emergence and advantages of machine learning-based IDS

-Objectives of the review

2-Remove Section 2.2: This section is unnecessary and can be omitted.

3-Remove Section 3: This section is also unnecessary.

4-Restructure the Paper: Given that the main focus of the paper is machine learning-based IDS, I suggest restructuring the paper. In Section 2, provide an overview of IDS in general, emphasizing the three different types of IDS (signature-based, anomaly-based, and hybrid-based).

5-Redundancy in Section 4: The content in Section 4 is already well-covered in existing literature. Instead of elaborating, you can cite relevant papers. Consider what new insights or added value your paper provides.

6-Systematic Literature Review: To enhance the robustness of your findings and minimize bias, consider conducting a systematic literature review.

 

Comments on the Quality of English Language

 Minor editing of English language required

Author Response

Thank you very much for your comments and constructive suggestions.
In the attached document we have provided a response to each of your comments.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The paper provides a detailed survey and analysis of various machine-learning techniques used to detect intrusions within IoT systems. It discusses the limitations of traditional intrusion detection systems and explores machine learning as a more adaptable solution to IoT environments. The paper covers a range of IoT security threats taxonomy and machine learning strategies including supervised, unsupervised, and deep learning methods, evaluating their effectiveness and practical applications in enhancing IoT security.

 

My major concern is that the focus of the paper appears to be somewhat unclear, as it allocates substantial portions to introducing a broad taxonomy of security threats and providing basic information on various machine learning models. They detract from a deeper exploration into the integration and specific application of machine learning techniques in IoT intrusion detection.

 

The detailed descriptions of basic knowledge, including machine learning models such as KNN, SVM, Decision Trees, ANN, and CNN, although informative, might be redundant for the target audience, who likely have a foundational understanding of these models. I suggest streamlining these sections to focus more on the application of these models—especially the more advanced models—specifically for IoT intrusion detection. The basic principles should be summarized briefly, with the focus shifted to their IoT applications, performance, and challenges.

 

The comparison between the surveyed works could be enhanced, like what the author did for the dataset comparison. Discussion should focus more on the unique challenges and requirements of applying machine learning models in IoT environments, with an emphasis on issues related to resource constraints, real-time processing needs, and the diversity of IoT devices. More detailed comparisons of their strengths and weaknesses with respect to IoT-specific metrics such as energy efficiency, processing time, and adaptability to various types of IoT attacks would also be helpful.

 

 

Why is section 7.1.3 on KNN in French?

Comments on the Quality of English Language

Some sections are not in English

Author Response

Thank you very much for your comments and constructive suggestions.
In the attached document we have provided a response to each of your comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

Dear All,

I have sent you a couple of comments to enhance your comprehensive study. But these comments were not taken into consideration.

Comments on the Quality of English Language

 Minor editing of English language required

Author Response

Thank you very much for your constructive comments.The version that was previously submitted does not allow you to view the changes made.However, all the corrections have been taken into account in the correction report submitted.We have also integrated a track changes in the article to better reflect all the corrections made.We hope that all these elements respond to your suggestions that are constructive to us. 

Reviewer 4 Report

Comments and Suggestions for Authors

I think the revision addresses some of my previous concerns, the paper is now in better shape. However, it seems that the author did not submit a version that highlights the changes made, which makes it difficult to track the modifications.

Additionally, the author should proofread the paper and correct typos. For instance, in section 6.1.2, 'SVms' should be corrected to 'SVMs'.

Author Response

Thank you very much for your constructive comments.The version that was previously submitted does not allow you to view the changes made.However, all the corrections have been taken into account in the correction report submitted.We have also integrated a track changes in the article to better reflect all the corrections made.We hope that all these elements respond to your suggestions that are constructive to us. 

 

Round 3

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors I appreciate the work done to edit the paper but my last comments were not considered.

6-Systematic Literature Review: To ensure the robustness of your findings and minimize bias, could you conduct a systematic literature review?

Comments on the Quality of English Language

Moderate editing of English language required

Author Response

Thank you for your valuable comments and suggestions regarding the importance of a systematic literature review. We greatly appreciate your concern to ensure the robustness of our findings and minimize bias.
We would like to inform you that for the completion of this work, we carried out a systematic review of the literature following a rigorous methodology. We defined criteria for the inclusion and exclusion of studies, focusing on articles published in peer-reviewed journals, international conferences, and technical reports relevant to IoT systems and machine learning attack detection. We searched recognized academic databases such as IEEE Xplore, PubMed, and Google Scholar, and developed a detailed search strategy using specific keywords to maximize coverage of relevant studies. We followed a selection process based on titles and abstracts, followed by a full reading of the selected articles to confirm their relevance. We also assessed the quality of the included studies to ensure that the results were based on solid, reliable research, and synthesized the findings.
Drawing on existing literature reviews in the field, we found that, for the most part, they were somewhat dated or did not include recent studies and research carried out in the field. Our review is intended as an update that offers a more comprehensive overview, providing information on current issues and challenges as well as emerging techniques. We hope this explanation clarifies our methodology and meets your expectations

Round 4

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

Still my comment where not handled specially the one related to the selection of the papers to be included in yuor literature review. 

As I have mentioned before you should show the reasons of selecting the paper and what are the bases of selecting them.

 

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Thank you very much for your constructive comments.
We have taken this into consideration and have introduced a section for "Materials and Methods" to explain our approach to selecting the articles used in this journal.
This has also enabled us to discover interesting articles such as the following 3 that we have mentioned in our document :

• Rafique, S.; Abdallah, A.; Musa, N.; et al. Machine Learning and Deep Learning Techniques 1294
  for Internet of Things Network Anomaly Detection-Current Research Trends. Sensors 2024, 1295
  24, 1968. https://doi.org/10.3390/s24061968. 1296
• Haque, S.; El-Moussa, F.; Komninos, N.; et al.. A Systematic Review of Data-Driven Attack 1297
  Detection Trends in IoT. Sensors 2023, 23, 7191. https://doi.org/10.3390/s23167191. 1298
•  Sarker, I. Machine Learning: Algorithms, Real-World Applications and Research Directions. SN 1299
  Computer Science 2021, 2, 160. https://doi.org/10.1007/s42979-021-00592-x. 1300

we hope that these final corrections meet your expectations to improve our document.

Round 5

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

I wolud like to thank you for the effort done to conduct a SLR.

For me this SLR is still not clear and you didn't mention the number of papers conducted at the begining and how you filtered them using inclusion/exclusion criteria and then after doing scaninng to the title and abstract and after doing the full text scan. Even you mention there as a lot of review appers done. But you didn't compare yuor results with others to show your contribution.

 

Comments on the Quality of English Language

Moderate editing of English language required

Author Response

Thank you very much for your comments. We did not previously integrate the elements of a systematic review since our basic job was to do a complete and comprehensive literature review. By following your instructions we have improved the Method section to cover and explain the selection criteria for studies and articles following the PRISMA standard. You will also find a PRISMA flow diagram.
We hope that these corrections respond to your suggestions in order to improve our document