Enhancing IoT Security in Vehicles: A Comprehensive Review of AI-Driven Solutions for Cyber-Threat Detection

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

R. Abreu et al.'s review paper examines how AI enhances IoT security, highlighting recent AI research in the field and including some discussion. However, I have the following concerns:

i) The introduction lacks depth. The authors rely on only four citations, offering limited support for their statements. A more thorough discussion with additional references to related work would be necessary.

ii) As a review paper, it is unusual for the authors to include a section on 'methodology.' I believe this detracts from the purpose of a review, which should focus on synthesizing existing research rather than describing methods.

iii) The related work section is presented as a list of studies without a clear logical structure. A more cohesive narrative or logical organization would help improve readability and comprehension.

Given these issues, I can not recommend the paper for publication at the current stage.  At least, the author should restructure the manuscript and add more citations.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript reviews the state-of-art studies regarding IoT security in vehicles. With the advance of AI technology, it's of great interest and importance to discuss its application in IoT security.

Overall, the manuscript is well-written, but it also has some major caveats that need clarifications and improvements.

1. Though the authors have been using "AI" throughout the manuscript, including the title. The main focus is actually machine learning and deep learning methods. However, AI is a much broader concept than machine learning and deep learning. Machine learning and deep learning are data-driven, but what's the research status in other types of AI algorithms? If the authors want to provide a comprehensive review in AI-driven solutions, then the discussion and literature review should be broader to include research other than machine learning and deep learning.

2. I don't think it's necessary or interesting to talk in length about what database you searched and how many articles you found or picked for this review (section 2). Readers are mostly interested in your own summary and understanding of the research field. When all IMPORTANT papers (either conference papers or from journals) are included and cited in your references, readers already has gained the information about database and number of articles.

3. The discussion section is well-written with some in-depth thoughts, but the conclusion section is not strong enough. Some statements are simply known truth (e.g., lines 665-666; lines 674 - 675 about the evaluation metrics) that may not provide interesting or new information to the readers.

 

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

With an emphasis on the automobile industry, this paper provides an extensive and methodical analysis of AI-driven security solutions for Internet of Things (IoT) devices. The writers assess popular datasets, examine a range of machine learning and deep learning approaches, and talk about the field's obstacles and potential paths forward. It is suggested actually this manuscript is well-structured and written, and pretty much really ready good for publication because of its rigorous methodology and insightful analysis; only a few small edits will improve its overall quality and impact. Here are the suggestions:

1(1) Personal comment is on the section 4.4, ajhtough the authors have given descriptions briefly, it might be better if could benefit from actually for a more specific mentioning of how these datasets, for example UNSW-NB15… and so on was preprocessed (e.g., whether it was balanced, or some details about feature engineering applied to them, what are the feature input, output features so on, and size?). This would really give readers a even better more clearer picture of how the AI models leverage the data. A visualized figure, or a table sorting this kind of thins out would be very good

(I(2) in the automotive-specific part, which is section 3.2, it might be a minor suggestions that actually the authors, they could try a bit effort for expand on the unique specific challenges of how we securing CAN bus networks against very complex attacks like bus-off attacks or targeted message injection and so on etc., which are basically not fully addressed by current AI-based solutions…

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The current manuscript might be good for publication.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript has improved a lot and I recommend publication.