Review Reports

Round 1

Reviewer 1 Report

This study, which is carried out within the scope of anomaly detection on microservice architectures that increase agility, scalability and usability in today's software architecture, is very interesting. I believe that the study with the following regulations can contribute more to the literature.

-Although the introduction is written well in general terms, there is a lack (clarity, fluency, etc.) especially in the part of the contribution of the study to the literature. The motivation and limitations sections are well written and increase readership. Maybe the types of attacks (APT attacks, malware etc.) encountered in microservice architectures can also be mentioned.

-By referring to more studies on these issues in the literature, you should better emphasize the difference between your work and the importance of the MLP method you used in your work.

-The method and infrastructure are explained very well in the study, congratulations. But to make the work even more valuable; It would be more appropriate to run the created dataset with different supervised machine learning algorithms (svm, knn, naive bayes etc.) and evaluate the results accordingly.

Author Response

The authors are not native English speakers, so the natural fluency might not be as good as one might desire. Nevertheless, the text was reviews according to the comments in order to address the general fluency and the particular aspects referred by the reviewers.

Author Response File: Author Response.pdf

Reviewer 2 Report

This article is interesting as it addresses a scenario for anomaly detection in microservices, mainly in a simulated environment with cloud services. I leave you some suggestions to improve the manuscript:

1.    Lines 181-197. I really don't see the need to add a source code, I think an algorithmic or pesudocode list could be better, where you can explain step by step the performance and functional anomalies.
2.    I find that Tables 1 and 2 create a bit of confusion on a first reading. For example, data on percentiles and features are read, but they are described below and it would be necessary to return to the table to corroborate what is tabulated. In this part you could start first describing which features were used as they were obtained and then refer to the tables.
3.    In section 2.5 Neural networks, the introduction to MLP is very short, I think you should go into more detail here to explain its structure, motivation and how it has been used in anomaly detection tasks.
4.    Section 3. Model and Results, specifically subsection 3.1. Data Preprocessing, explains without much detail what the purpose of preprocessing is, for example why use  MinMaxScaler and not others?
5.    Sub-section 3.2. Hyperparameter Tuning, describes the purpose of improving the natural configuration values of the MPL, but does not talk about what ranges of values were used. I suggest referring to the following manuscript for further discussion of this topic:

Gonzalez-Cuautle, D., Hernandez-Suarez, A., Sanchez-Perez, G., Toscano-Medina, L. K., Portillo-Portillo, J., Olivares-Mercado, J., ... & Sandoval-Orozco, A. L. (2020). Synthetic minority oversampling technique for optimizing classification tasks in botnet and intrusion-detection-system datasets. Applied Sciences, 10(3), 794.

6.    In the Discussion section, I find it necessary to look for more literature detailing why MLP is a good algorithm for anomaly detection tasks specific to microservices. For example, decision trees are prone to overfiting while MLP can be customized towards a more efficient objective function...

7.    The section named Conclusions and Future Work talks about practical implications of implementing the model, but what are they? could the model be suitable for an incremental or online learning scenario?

Remember these are just suggestions to improve the manuscript as you are the experts on the subject.

Regards

Author Response

The authors are not native English speakers, so the natural fluency might not be as good as one might desire.

Nevertheless, the text was reviews according to the comments in order to address the general fluency and the particular aspects referred by the reviewers.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Dear authors, thank you for attending to the suggestions provided in the first revision. I consider that the new contributions give a good traceability to the reading, understanding and contribution to the state-of-the-art of this manuscript. I conclude that the document can be considered for publication.