Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper presents an approach to enhancing identity verification in e-KYC using super-resolution, CNNs, and Bayesian uncertainty estimation.

Following improvements are suggested.

 

1. Abstract – at the end include the obtained highest performance values
2. Introduction – clearly mention the research questions address by this study. And in the discussion section, justify how you achieved those RQs with your methodology and results.
3. Introduction - state the limitations of traditional methods beyond just susceptibility to fraud. Cite relevant statistics.
4. When you cite the related studies, do not include all author names. Follow standard citing practices. For example, instead of “Pei Li, Loreto Prieto, Domingo Mery, and Patrick J. Flynn [12]” – you can state as “Li et al. [12]”
5. Related work – instead of describing set of paper summaries, analyse them, establish links between related studies.
6. Describe more latest related work (say 2024, 2025) in facial recognition for identity verification, super-resolution techniques applied to facial images, and the use of uncertainty estimation in deep learning for computer vision.

 

7. Justify the methodology design with the integration of Super-Resolution preprocessing, CNN, and Monte Carlo Dropout. Explain how they complement each other for the task of reliable identity verification.

 

8. Provide more details about the architecture of the CNN used. What specific CNN model was employed? What were the key layers and parameters? If it is customized CNN, give details. Is it the later defined Deep Recursive Residual Network (DRRN) architecture? (explained in section 3.1?)  - if so, mention that in figure 1.

1. The choice of resizing images to 24×24 pixels for training and subsequent use of super-resolution should be justified with references or ablation studies. Why was this resolution chosen, and how does it compare to standard benchmarks in facial recognition?

10. Describe the dataset(s) used for training and evaluation in detail. Include information about the size, diversity, and any preprocessing steps applied. If the dataset contains low-resolution images, specify how these were obtained or simulated.

1. Consider including a comparison with other state-of-the-art methods, both with and without uncertainty estimation, to contextualize the performance gains.

12. It would be better to extend table 8 to compare the performance of the proposed framework with baseline methods or existing state-of-the-art techniques in 2024, 2025
13. Discussion - Explain why the proposed approach performs as it does. Discuss the impact of each component (Super-Resolution, CNN, uncertainty estimation) on the overall performance.

1. Discussion – clearly state the novel contribution compared to the LATEST (say 2024, 2025) existing work.  Please elaborate on how the proposed framework significantly advances the state of the art compared to recent literature, especially regarding practical deployment and compliance with standards.
2. Discuss, with the use of the Monte Carlo Dropout and its integration with CNN, how uncertainty is quantified and how it influences the final decision-making process.
3. The manuscript mentions both epistemic and aleatoric uncertainty, but the methods for distinguishing and quantifying these are not sufficiently detailed. Please clarify how each type of uncertainty is measured and utilized within the system.
4. The robustness of the model to various degradations (noise, blur) is mentioned, but quantitative results (tables/figures) comparing performance under different conditions would strengthen the claims.
5. Explain how the proposed framework meets alignment with international standards (IAL, AAL). Are there any certification or compliance tests performed?
6. Discuss potential challenges and limitations in deploying the proposed system in real-world e-KYC settings, such as computational requirements, latency, and integration with existing infrastructure.

 

Author Response

We sincerely thank Reviewer 1 for the valuable comments and constructive suggestions. We have carefully addressed all the points raised, and the detailed responses can be found in the attached document. All revisions have been incorporated accordingly to improve the quality and clarity of the manuscript.

Please kindly refer to the attached file for our point-by-point responses.

Thank you for your kind consideration.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

• The article contains numerous mathematical equations. Please confirm whether all the equations used have been defined by the authors. If not, kindly provide appropriate references for those derived or adapted from prior work.

• Perform a timing analysis of the algorithms utilized in the article to evaluate their computational efficiency.

• The article employs cosine similarity to measure similarity. Before selecting this method, did you evaluate or compare other similarity measures or algorithms? If so, please provide details and justify the final choice.

• Justify the statement: “The identification process is closely tied to registration protocols and involves stringent verification steps that emphasize document authenticity and image-based identity matching.” Explain the methods or techniques used to ensure these verification standards.

• Provide a proper reference for the dataset used in the study. Also, specify the size of the dataset in megabytes (MB) or gigabytes (GB).

 

 

 

Author Response

We sincerely thank Reviewer 2 for the valuable comments and constructive suggestions. We have carefully addressed all the points raised, and the detailed responses can be found in the attached document. All revisions have been incorporated accordingly to improve the quality and clarity of the manuscript.

Please kindly refer to the attached file for our point-by-point responses.

Thank you for your kind consideration.

Author Response File: Author Response.pdf