All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Since all of the reviewer's suggested changes have been implemented in this latest version, the paper is accepted. Congratulations.
[# PeerJ Staff Note - this decision was reviewed and approved by Sedat Akleylek, a PeerJ Section Editor covering this Section #]
Authors have carried out my suggested changes. Hence, Manuscript is accepted
Authors have carried out my suggested changes. Hence, Manuscript is accepted
Authors have carried out my suggested changes. Hence, Manuscript is accepted
Reviewer 1 has pinpointed several issues, such as:
- The research gap discussion is broad and needs more concrete examples of existing limitations.
- The Literature Surveylacks of critical analysis; i.e without in depth discussion of the related work's drawbacks and misses a comparison with the proposed system.
- the theory seems to be disconnected: there is lack of explanation of how the theoretical framework directly contributes to the proposed method and its major contributions
- in the Experimental Design, there is insufficient discussion of statistical significance or confidence in the results.
- about the findings, there is also lack of discussion on the results which are obtained.
- paper needs a better grounding for the authors' claim that higher performance should be provided
So, please work so to address all of Reviewer 1 comments and critiques.
Abstract
o Overuse of jargon might make it less accessible to general readers. Provide qualitative analysis
o Claims of significant improvement lack direct reference to specific metrics or benchmarks.
Introduction
o Repeated information about an AI explainable intrusion detection system, which could be streamlined.
o The research gap discussion is broad and needs more concrete examples of existing limitations.
Literature Survey
o Lack of critical analysis; it mostly lists prior works without discussing their drawbacks in depth.
o Missing comparison with the proposed system to highlight its unique contributions.
Theoretical Framework
o Heavy reliance on prior studies without explaining how this framework directly contributes to the proposed method and its major contributions
Proposed System
o Equations and algorithms are not always intuitive; more illustrative examples are needed.
o Figures and diagrams are not adequately referenced in the text, reducing the clarity of the explanation.
o Clarity of algorithms needed to be improved
Experimental Setup and Results
o Results are only compared with three other protocols, which limits the scope of evaluation.
o Insufficient discussion of statistical significance or confidence in the results.
o Figures are presented without proper context or explanation in the text on why the proposed system always performs better than existing
Performance Assessment
o Lack of discussion on the results which are obtained.
o On what basics does the author claim higher performance should be provided clearly
The references are outdated. Authors are suggested to include the following latest papers if they are relevant.
1. An enhanced whale optimizer-based feature selection technique with an effective ensemble classifier for a network intrusion detection system
2. An improved Harris Hawksoptimizer-based feature selection technique with an effective two-staged classifier for a network intrusion detection system
3. Intrusion detection system and fuzzy ant colony optimization-based secured routing in wireless sensor networks
4. Intrusion detection system extended CNN and artificial bee colony optimization in wireless sensor networks
5. Secure and optimized intrusion detection scheme using LSTM-MAC principles for underwater wireless sensor networks
6. Intelligent IDS in wireless sensor networks using a deep fuzzy convolutional neural network
The manuscript underwent professional language editing, addressing prior concerns about awkward phrasing and grammatical issues. Sentence construction in the abstract and methodology is now clearer and more concise.
The authors have significantly enriched the literature discussion. They contrasted their work with recent GAN- and optimization-based IDS approaches and provided a more analytical narrative. Citations are now more accurately distinguished between biological and algorithmic references.
Figures and tables are now better described, including dataset context, metric types, and classification tasks.
The manuscript now includes a well-structured ablation study and a clarified hypothesis. The logical integration of IMOA, WGAN-GP
The paper has included a more detailed explanation of the IMOA algorithm's convergence behavior, complexity comparisons, and literature-supported justification for attention mechanisms.
The paper clearly defines its objectives and how it addresses real gaps in existing IDS frameworks. Contributions are now restructured under “Novelty,” “Advantage,” and “Practical Benefits,” providing sharper articulation.
An Ethical Compliance Statement has been included. A detailed computational complexity and runtime comparison further supports the practical applicability of the approach. Results are reported with 5-fold cross-validation and supported with significance testing.
All methods are described in sufficient technical depth. The updated GitHub repository and clear explanation of preprocessing and training steps ensure reproducibility. Architecture and parameter settings for each model are fully documented.
A dedicated Replication Rationale section has been introduced, clarifying how the study replicates and extends previous ensemble and GAN-based IDS models. Comparative benchmarking across three datasets confirms the value of the hybrid approach.
Conclusions are now presented with appropriate scope, avoiding overgeneralization.
The authors have responded to reviewer concerns with extensive, well-structured, and scientifically sound revisions.
**PeerJ Staff Note:** Please ensure that all review and editorial comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
**Language Note:** The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title). Alternatively, you should make your own arrangements to improve the language quality and provide details in your response letter. – PeerJ Staff
The authors should consider the following suggestions provided by the reviewer to improve the scientific depth of their manuscript, as well as they should also address the following comments to improve the quality of the presentation of their manuscript:
1- The author(s) have to further improve the technical writing of the paper. This manuscript needs careful editing by someone with expertise in technical English editing, paying particular attention to English grammar, spelling, sentence structure, typos, spaces, and indentation errors so that the goals and results of the study are clear to the reader.
2- In the Introduction part, the strong points of this proposed article should be further stated. Not clear what the originality of this work is? What makes it different from existing work?
3- The major contribution of the research needs to be properly described.
4- The quality of all figures is poor; the author(s) must redraw them with high quality. Some text in figures is difficult to read.
5- Abbreviations used need to be elongated only once, at their first usage in the manuscript. Authors are advised to correct them.
6- It is preferable to place figures/tables directly after referring to them in the body of the text.
7- More comprehensive evaluations are needed for publication.
8- The authors are encouraged to add future works to the conclusions section.
9- Some of the references are out of date. This may be an indication that the research is old or that this field of research has reached its end and is no longer an interesting topic. Provide more up-to-date references (in SCOPUS, the papers published within 3 years are used to calculate CiteScore). Replace old references with new references, preferably between 2022-24.
-
-
Basic Reporting
1. Clear and Unambiguous, Professional English Used Throughout
• The manuscript is largely written in professional and technical English.
• Technical terms are used appropriately, and the flow is generally coherent.
• Some grammatical issues, such as verb agreement, and inconsistent article usage.
• Sentence construction is occasionally awkward or overly verbose, especially in the abstract and methodology sections.
• A full language editing pass by a native or professional editor is recommended to improve readability and clarity.
2. Literature References, Sufficient Field Background/Context Provided
• The introduction thoroughly explains the challenges in intrusion detection systems (IDS), such as high-dimensional data and class imbalance.
• The paper references several foundational and recent works, giving context to the proposed approach.
• Some citations (e.g., [14], [15]) are repeated frequently and are used to support both technical and behavioral aspects without distinction between primary biological sources and algorithm design references.
• The literature review in the “Literature Survey” section is somewhat superficial. Though summarized in a table, the narrative explanation is limited.
• Expand the discussion in the literature review to include more critical comparisons of similar GAN-based or optimization-based IDS models.
• Clarify the novelty more clearly by contrasting with 2–3 recent state-of-the-art papers.
3. Professional Article Structure, Figures, Tables. Raw Data Shared
• The structure of the article is comprehensive and well-aligned with standard research formats (Abstract, Introduction, Methods, Results, Discussion, Conclusion).
• Visualizations include loss curves, accuracy plots, PR and ROC curves, and confusion matrices.
• While figures and tables are informative, some captions lack sufficient detail to understand them independently.
• The article references the availability of source code on GitHub, but it is unclear whether the actual raw data used in the experiments (e.g., processed/augmented datasets) are publicly available.
• Improve figure captions for clarity (e.g., specify datasets, metrics, and classification type).
• Provide direct links to downloadable datasets or processed subsets used.
• Include a Data Availability Statement per journal policy.
4. Self-contained with Relevant Results to Hypotheses
• The proposed framework is clearly defined with a logical sequence of components: IMOA for feature selection, WGAN-GP for data augmentation, and a CNN/LSTM/FNN ensemble with attention.
• The experiments are conducted on three benchmark datasets and evaluated on both binary and multiclass tasks.
• The motivation behind choosing the Indian millipede as a biological model is extensive but needs clearer justification for its algorithmic advantages over existing metaheuristics.
• No ablation study is conducted to demonstrate the individual contribution of IMOA, WGAN-GP, or the attention mechanism.
• Consider including an ablation study to isolate and validate the contributions of each component in the pipeline.
• Clarify the research hypothesis more explicitly in the Introduction and connect it directly with the evaluation metrics in Results.
5. Formal Results Should Include Clear Definitions of All Terms and Theorems, and Detailed Proofs
• The mathematical modeling of the Indian Millipede Optimization Algorithm (IMOA) is detailed and logically structured.
• Equations are provided for fitness evaluation, update rules, penalty terms, and attention scores.
• While the IMOA is mathematically defined, theoretical guarantees (e.g., convergence or time complexity) are not discussed.
• The attention mechanism is well-documented mathematically but lacks formal analysis or theoretical justification.
• Include more discussion or references regarding the theoretical properties (e.g., convergence behavior) of IMOA.
• Consider discussing computational complexity and runtime comparisons with similar algorithms.
Experimental Design
1. Original Primary Research within Aims and Scope of the Journal
The article fits squarely within the scope of PeerJ Computer Science, particularly in the domains of artificial intelligence, cybersecurity, optimization algorithms, and deep learning applications.
The manuscript presents original primary research that combines a new bio-inspired optimization algorithm (IMOA), a Wasserstein GAN-GP for data augmentation, and a dynamic attention-based deep ensemble model for network intrusion detection.
2. Research Question Well Defined, Relevant & Meaningful
• The problem statement is clear: traditional IDS face limitations in handling high-dimensional, imbalanced datasets and identifying novel attack types.
• The research question is how a hybrid AI framework combining optimization (IMOA), generative modeling (WGAN-GP), and deep learning (attention-based ensemble) can address these challenges.
• The contribution is positioned as novel by introducing IMOA (inspired by Indian millipede behavior), applying WGAN-GP to improve minority class detection, and using dynamic attention to improve model fusion.
• A solid knowledge gap is articulated — existing systems either fail on imbalanced datasets or cannot dynamically adapt model weighting. This work aims to fill that gap.
• While the research contributions are listed (Section 1.5), they could be more sharply contrasted against existing works. What specifically makes IMOA better than other metaheuristics? How does dynamic attention compare with fixed-weight ensembles in IDS?
• Strengthen the critical review of prior optimization methods (e.g., PSO, GA, ACO) and ensemble strategies, showing why IMOA and dynamic attention are advantageous.
• Consider rephrasing the contributions more clearly in terms of “novelty,” “advantage over prior work,” and “practical benefit.”
3. Rigorous Investigation Performed to a High Technical & Ethical Standard
• The experimental evaluation is thorough, conducted on three benchmark datasets: UNSW-NB15, CIC-IDS2017, and H23Q. These datasets are diverse, recent, and widely used in IDS research.
• The study includes a broad range of evaluation metrics (accuracy, precision, recall, F1-score, confusion matrices, loss functions, PR curves, ROC curves, attention weight plots).
• Multiple classification types are analyzed (binary and multi-class), providing robustness.
• There is no explicit statement of ethical compliance (e.g., license or use terms of datasets, confirmation that no private data is included). While these datasets are public, this should be declared.
• The runtime complexity or computational resource usage of the model is not benchmarked or compared against baselines, which limits assessment of practical applicability.
• Add an Ethical Compliance statement, confirming all datasets are publicly available and appropriately used.
• Provide a computational complexity analysis or runtime comparison to demonstrate scalability, particularly since the model includes GANs, three DNNs, and an optimization loop.
• Include baseline comparisons for computation (e.g., runtime, memory usage, number of parameters) against conventional methods or simpler ensembles.
4. Methods Described with Sufficient Detail & Information to Replicate
• The paper provides pseudocode and mathematical formulations of the IMOA algorithm in detail (equations 1–15), covering its biologically inspired behaviors.
• The architecture and parameter details for CNN, LSTM, FNN, and WGAN-GP are well-described.
• The data preprocessing pipeline is clearly explained, including cleaning, transformation, scaling, and augmentation steps.
• The code repository is available on GitHub, which promotes reproducibility:
Validity of the Findings
1. Impact and Novelty Not Assessed (Replication Acceptable if Justified)
• The study is not judged on novelty, but the authors introduce a new Indian Millipede Optimization Algorithm (IMOA) and apply WGAN-GP for data balancing, which in itself is a meaningful replication and integration of several known AI techniques.
• The use of benchmark IDS datasets (UNSW-NB15, CIC-IDS2017, and H23Q) provides a platform where performance comparisons and replication are not only meaningful but also necessary for progress in the field.
• Although the paper is innovative in combining multiple techniques, it lacks a replication rationale in the traditional sense (e.g., validation of an existing method under new conditions).
• It does not discuss how this work serves as a comparative validation against existing ensemble IDS frameworks in terms of reproducibility or computational resource usage.
• Explicitly frame part of the study as a replication or validation of established IDS methods, especially the ensemble or GAN-based techniques. This would justify inclusion under PeerJ’s guidelines.
• Add a section comparing computational performance (e.g., training time, memory use) with existing IDS approaches for better grounding in replication/benchmarking.
2. Underlying Data Provided; Robust, Statistically Sound & Controlled
• The paper evaluates the proposed model on three widely accepted, publicly available benchmark datasets.
• Data preprocessing steps, such as cleaning, feature selection (via IMOA), and normalization are well documented.
• Performance is evaluated using standard and multiple metrics (accuracy, precision, recall, F1-score, PR/ROC curves, confusion matrices), providing robust quantitative support.
• While performance is analyzed in detail, there is no mention of data variability control such as:
o Cross-validation
o Multiple runs with different seeds
o Reporting variance or standard deviation
• No statistical significance testing is performed (e.g., t-tests, confidence intervals) to support comparative claims.
• Include results over multiple runs (e.g., 5-fold or 10-fold cross-validation) and report mean ± standard deviation.
• Perform statistical significance testing where appropriate to confirm claims (e.g., that IMOA outperforms other optimizers).
• Add information about random seed control, stratified splitting, and potential data leakage prevention measures to ensure validity.
3. Conclusions Well Stated, Linked to Original Research Question & Supported by Results
• The conclusions are directly tied to the main research question: can an IDS that uses IMOA, WGAN-GP, and attention-based ensemble methods outperform traditional approaches?
• The results convincingly support this in terms of classification metrics on both binary and multiclass tasks.
• Some conclusions could be perceived as overly strong or generalized (e.g., the model is “superior” without statistical support or runtime comparison).
• The conclusions would be stronger with clear limitations outlined and future directions discussed.
• Reframe some conclusive statements to reflect the scope of findings (e.g., “in our experiments,” “on benchmark datasets,” rather than universally superior).
• Add a Limitations section highlighting:
o Dependency on hyperparameter tuning
o Generalization to real-time systems not evaluated
o Limited diversity in types of optimization algorithms compared
• Suggest future research avenues such as real-time deployment, lightweight model variants, or testing on encrypted traffic.
The article presents an interesting and innovative approach by integrating a novel bio-inspired optimization algorithm with deep learning and GANs for intrusion detection. However, substantial improvements in language, experimental validation, and theoretical justification are required before it meets the standards for publication.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.