Hybrid Phishing Detection Based on Automated Feature Selection Using the Chaotic Dragonfly Algorithm

Round 1

Reviewer 1 Report

The paper deals with Hybrid Phishing Detection Based on Automated Feature Selec- 2 tion Using the Chaotic Dragonfly Algorithm. The article has some new scientific assets, but to be published, the paper must be enhanced in state of the art section, a better explanation of the model, better evaluation a better presentation of results. After major revision, I recommend accepting the article.

Please use proofreading.

Author Response

The paper has been modified and proofread.

Author Response File: Author Response.pdf

Reviewer 2 Report

Summary

With the growing prevalence of phishing attacks, researchers have become increasingly interested in network security. The data generated often contain irrelevant and inappropriate features that can impact the accuracy of machine learning and classification algorithms; there is a need for a robust method to effectively detect phishing threats and enhance detection accuracy.

The paper proposes a novel approach that combines the Chaotic Dragonfly Algorithm for feature selection with KNN and SVM classification algorithms. The results show that this approach improves classification accuracy compared to using the same methods without feature selection. The robustness of the proposed design is also evaluated using three public datasets.

Strong points:

One of the paper's strengths is the innovative use of the Chaotic Dragonfly Algorithm for feature selection. This algorithm provides an effective way to select relevant features for classification. The experimental results demonstrate improved accuracy, highlighting the efficacy of this approach.

Another strong point is the comprehensive evaluation using multiple classification algorithms and three public datasets. This helps establish the robustness and generalizability of the proposed design. By considering different algorithms and datasets, the authors provide a complete understanding of the approach's performance and its potential applicability to various domains.

Weak Points:

However, there are still some weak points to be improved.

1. Figures: Ensure that the figures in the paper are of high resolution to enhance readability and clarity. This will help readers better understand the presented information.
2. Performance metrics: While accuracy and standard deviation are commonly used metrics, consider including additional performance metrics such as F1-score, Recall, and Precision. This will provide a more comprehensive evaluation of the proposed approach's effectiveness.
3. Table index: Review the table index to ensure it is correctly aligned. For example, check lines 218 and 228 to ensure they correspond to the correct tables.
4. Parameter values: Justify the selection of parameter values for the models used in the experiments. Explain the reasoning behind choosing specific parameter values, as this will enhance the reproducibility and reliability of the results. For instance, clarify the rationale behind the parameter values in line 223.
5. Abbreviations: Address the issue of overlapping abbreviations. In line 45, Dragonfly Algorithm (DA) and data augmentation (DA) are both abbreviated as "DA," which may cause confusion. Consider using typical abbreviations or rephrasing the text to avoid ambiguity.
6. Limitations: Clearly identify and discuss the limitations of the proposed approach in the paper. While line 347 mentions limitations, they are not described in detail. Expand on the limitations, explaining potential constraints or drawbacks of the proposed method and suggesting areas for future improvement.

Overall Review:

The paper presents an interesting approach that combines the Chaotic Dragonfly Algorithm for feature selection with classification algorithms. The results show improved accuracy compared to using the same classification methods without feature selection. The comprehensive evaluation on multiple datasets adds credibility to the proposed design. However, the paper would benefit from addressing the identified weak points, such as including additional performance metrics and providing a more detailed discussion of limitations.

Considering these factors, I believe the paper has the potential to make a contribution to the journal With the necessary revisions and improvements, it can be a valuable addition to the existing literature on feature selection and classification algorithms. I recommend that the authors carefully address the provided feedback to strengthen their paper and enhance its overall quality before the paper is accepted.

The paper can benefit from a native reader's review. Not that is ungrammatical, but some sentences sound like written by inexperienced researchers.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper proposed a robust method based chaotic dragonfly algorithm to detect phishing threats. The proposed scheme provides more accurate results than all other baseline classifiers. In general, this paper is easy to follow and in a very important area. However, there are some problems which must be solved before it is considered for publication, once resolved, would significantly improve the paper:

 

1) The contributions of the paper are inflated. They could be easily reduced to three.

 

2) Related work needs a few sentences to summarize the research status of the current work as a whole. The authors just simply list the basic idea of the existing schemes and the strength and weakness. The targeted problem addressed by the method in this paper is not proposed.

 

3)Materials and Methods introduces too much technical background, and the research methods are not explained sufficiently.

- Figure 2 needs to clarify the input and output of the algorithm.

- Eqs (9.1) to (9.8) did not appear in the paper.

- Every component in the system framework needs a piece of text to describe its functionality.

 

4) This paper is lack of sufficient explanation of the experiment results. You need to explain the results of the feature selection method based Dragonfly algorithm in detail and analyze and discuss the rationality and validity of the selected subset of features. Accuracy alone is not enough.

There is a lack of verification of the robustness of this method.

 

5) Figure 2 are a bit blurry. Please consider replacing it with clearer ones.

Authors need to recheck the paper for grammatical and formatting errors, e.g., there are two Equations (1) in the paper.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

According to the highlighted changes, the authors did not add to the article everything that the opponents wrote.

English is ok

Author Response

The paper was modified according to the reviewer's comments. Hence, lines 84-99 and lines 158-174 were added to the related work section. In addition, a new section 3.3 were added in the Materials and Methods. In the results section, a clear explanation was added from lines 281-302. Finally, two references were added.

Author Response File: Author Response.pdf