An Adversarial Example Generation Algorithm Based on DE-C&W

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes a novel adversarial example generation algorithm based on improved differential evolution and C&W attack. This algorithm enhances attack effectiveness by optimizing loss functions. It also reduces query costs and improves transferability. Finally, it strengthens the security of deep learning models against adversarial attacks.

Overall good paper. I would like to add few points as suggestion.

1. What specific consensus mechanism is used in the proposed framework?

2. How does the system handle robustness in practical applications, such as real-time systems?

3. Adversarial examples generated for a specific model may not effectively transfer to other models. How this study will handle this?
4. How the selection process is handled based on fitness values in algorithm for adversarial examples ?
5. Explain more about real-world scenarios related to this study.

6. Provide a structured comparative table summarizing existing studies and highlighting how your work differs.

7. Please provide detailed adaptive strategies for adjusting these parameters based on real-time performance feedback during model training.

8. Provide additional metrics like standard deviation of accuracy rates to show performance of model's stability. 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

please check the word file

Comments for author File: Comments.pdf

Comments on the Quality of English Language

English is ok but need to correct some typos errors

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Congratulations, it's a good article.

Some improvements:

#1 - Section 3.1 needs some improvement.
#2 - Section 3.2 needs some improvement in text and formulas.
#3 - Section 3.3 needs some improvement.
#4 - Some references to Figures do not maintain a coherent notation.
#5 - Figure 4, 5 and 6 should be improved.
(see pdf in annex).

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Well written manuscript, the message is clear and method is logical, and enough details are provided. I like how the evolution algorithm was utilized in the proposed approach, and the results validate the proposed approach. I actually don't have constructive feedback. Two minor comments, the resolution of Fig.5 needs to be improved, it is impossible to see the difference between the methods or the added perturbations; also, please reference C&W early in the manuscript, the term was used multiple times before actually being referenced.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf