CBA-CLSVE: A Class-Level Soft-Voting Ensemble Based on the Chaos Bat Algorithm for Intrusion Detection
Round 1
Reviewer 1 Report
The paper presents a soft voting ensemble which increase intrusion detection performances.
The paper is well written, many comparisons being provided in order to prove better performances, compared with other methods.
However, few questions arise regarding some aspects.
Related works for class level soft voting method based on the chaos bat algorithm are missing; only related work for soft voting methods are presented in the paper.
For equation (2), no measurable information about possible values (or range) for the weight of classifier (wi) is provided. In line 184 the authors mention that a large weight means a stronger classification performance. What means large?
Also, in chapter 3.3 The Basic Bat Algorithm, some references must be provided; the information regarding clasic bat algorithm is not new.
The expressions for chaper 5.1 Evaluation (eq. 16-21) are not new, references must be added.
Why the overall performance of the base learners without feature selection on NSL-KDD dataset is better than overall performance obtained with feature selection? For example, the best accuracy in Table 5 is 98.45 and in Table 8 is 97.50. The feature selection is a drawback of the proposed algorithm?
Minor spelling errors
Misuse detection in 36 shoul be ”misuse detection”
chaos initialization in line 253 should be ”Chaos initialization”
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This paper proposes CBA-CLSVE, which is a soft ensemble model based on Chaos Bat Algorithm, SVM, K-NN, and RF for intrusion detection purposes.
The paper flows well and is easy to read. The authors did a good job of presenting the background leading to their contributions.
I have a few suggestions that the authors should take into consideration:
- All abbreviations should be revised and the first letter of each word should be capitalized. For example, Chaos Bat Algorithm.
- Did the authors use the CICIDS2017 main attack classes or sub-classes? It is recommended to clarify this.
- In section 5.3, isn't this 10-fold cross-validation, why didn't the authors use the scientific term?
- In Table 12. the papers are dated 2018 and 2020. It is advised to compare with more recent work.
- In Table 2, it is advised to mention how the optimal parameters are reached.
- English grammar should be thoroughly revised.
- In section 2, it is recommended to add recent papers that proposed ML and DL models for IDS. The authors only presented ensemble models from the literature.
- The list of contributions should be added at the end of the introduction section as well as the organization of the paper.
- Section 4 should be at the beginning of page 6.
- In Figure 1, please review the sentences and their consistency in terms of wording and capitalization.
- Will the authors make their code available for reproducibility?
Author Response
Please see the attachment.
Author Response File: Author Response.docx
