A Heterogeneous Machine Learning Ensemble Framework for Malicious Webpage Detection

Round 1

Reviewer 1 Report

Referee report

This study proposes an ensemble approach for detecting malicious webpages among many adversary activities. In recent years, detecting malicious web links has become a vital security issue. An ensemble approach using different machine learning models yields better detection performance than the existing single model according to the information on the computational experience that is included in the paper. Therefore, I think that the paper makes a contribution and has the potential to be published in the Applied Sciences. However, I summarize in the GENERAL COMMENTS as follows:

 

GENERAL COMMENTS

1. The novelty of this manuscript is limited. The authors should clearly explain the importance and value of this study. In order to point out the technical contribution of the paper, please summarize the technical achievements of this work.

2. The authors propose an ensemble approach for detecting malicious webpages among many adversary activities. But, the description of the proposed approach is somewhat simple. Please provide approach principles in detail. Please highlight the key research of the proposed approach clearly.

3. The results are not significant enough for journal publication yet. More comprehensive evaluations are needed for journal publication. The authors should compare the proposed method to other ensemble approaches.

4. The reference format is incorrect. Please ensure that all references in the Reference list are in compliance the reference format of Applied Sciences. Please recheck and revise all references according to the reference format of Applied Sciences.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper present a new method for detecting malicious webpages among many adversary activities. .

The research idea is interesting but some considerations should be noted.

1. Highlight more the contributions of the work in the abstract and in the introduction.

2. In the Introduction, a literature review should be carried out comparing other similar works. The work has few references.

3. Figures should be improved, they are of low quality and very poor resolution

5. A discussion of the results should be presented, showing the advantages and disadvantages of the opposite method.

6. Future works should be better highlighted at the conclusion, also because the work is a first step in the process.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Reviewer comments:

1.     The overall performance of different machine learning models varies depending on data features, and ensemble approach also depend upon the data features then how this technique will improve the performance of ML model?

2.     Include the ROC Curve analysis of given data.

3.     From the Table 3, why the accuracy of model is high for Random forest and XG Boost and less for SVM and CNN. Give scientific explanation in-depth with reference paper results.

4.     What basis the weighted values are added in the ensemble method.

5.     Any machine learning technique will predict the output results by their statistical way of algorithm according to available datasets. What is the novelty of this work? How differ from others work particularly data analysis.

6.      When a reputation score is 8 or more, all predictions are true positive. How to achieve this?

7.     Include recent literature survey of last 2 years.

 

 

Author Response

"Please see the attachment

Author Response File: Author Response.docx

Reviewer 4 Report

This manuscript proposed a modified machine learning ensemble framework for malicious webpage detection, the research topic looks interesting. My comments are as follows:

1) English writing improvement is suggested.

2) Both motivations and contributions are unclear.

3) It commonly uses “Related Work” as the title of Section 2, not “Related Works”.

4) More comments about the existing works are suggested in Section 2.

5) High-quality figures are strongly suggested.

6) More evaluation metrics should be considered in the experiments, such as G-mean.

7) More state-of-the-arts machine/deep learning should be included in the experiments to make the experiments more sufficient or at least discussed in this paper. Some papers are recommended: multi-objective optimization-based adaptive class-specific cost extreme learning machine for imbalanced classification, Neurocomputing. Non-iterative and fast deep learning: Multilayer extreme learning machines, JFI.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

This study proposes an ensemble approach for detecting malicious webpages among many adversary activities. In recent years, detecting malicious web links has become a vital security issue. An ensemble approach using different machine learning models yields better detection performance than the existing single model. The authors carefully revised the manuscript and made some changes to the version according to the comments of the reviews. Therefore, I think that the paper makes a contribution and has the potential to be published in the Applied Sciences.

Author Response

Dear Reviewer,

 

Thank you for reviewing our paper.
Thanks for your valuable comments, my thesis has been improved.

We will continue to contribute to the Cyber Security through good research.

Thank you

 

Sincerely, Sung-Sam, Hong PhD. Adjunct Professor Jangan University

Reviewer 4 Report

The authors did not address all my issues well.

Author Response

"Please see the attachment."

Author Response File: Author Response.docx