Internet-of-Things Traffic Analysis and Device Identification Based on Two-Stage Clustering in Smart Home Environments

Round 1

Reviewer 1 Report

The author proposes an IoT traffic analysis and device identification method based on two-stage clustering for smart home environments. I only have a few observations:

1. The obtained accuracy (86.9%) should be compared with the related works.

2. The authors talk a lot about security... but I didn't understand the connection between the proposed method and device security. Here it should be detailed more.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper proposes a new method for Internet of Things (IoT) traffic analysis and device identification in smart home environments using a two-stage clustering approach.

Objective: The aim is to distribute the computational costs of traffic analysis, understand how IoT devices operate for safety, and identify devices based on traffic patterns over time.

Background Issues:

• IoT device identification models need frequent updating as devices change in homes. Most methods don't consider the computational load for this.
• Processing large volumes of IoT traffic data in one place causes heavy system loads. Sending all data to the cloud also raises privacy issues.
• Existing methods identify devices based on traffic in fixed time windows, overlooking variations in device communication cycles.

Proposed Method:

• Two-stage clustering of traffic data
  • Stage 1 clusters data at local home gateways
  • Stage 2 clusters features from stage 1 on a cloud server
  • Reduces bias and communicates less raw data to cloud
• Transforms traffic into numeric time series based on patterns
  • Represents device behavior over time
  • Used as features to identify devices

Experiments & Results:

• Compared two-stage clustering performance to a centralized one-stage method
• Showed two-stage method extracts more detailed traffic patterns
• IoT device identification accuracy of 86.9% using LSTM network on time series features
• Significantly outperformed one-stage method in accuracy (61.9%)

Conclusions:

• IoT traffic analysis via two-stage clustering effectively captures features and behavior
• Enables distributed processing to reduce computation and communication loads
• Time series representations accurately identify IoT devices based on traffic patterns
• Proposed system is promising for deployment in real smart home environments

Future Work:

• Improve identification accuracy
• Implement and test a real-world prototype
• Evaluate calculation and communication costs
•  
• 
  

  While the paper presents an interesting approach to IoT traffic analysis and device identification, there are some areas where the state of the art is not adequately detailed, and the research design lacks explicit statements regarding relevance and rigor.

  The paper mentions that the proposed two-stage clustering method outperforms a centralized one-stage method in terms of accuracy. However, it lacks a comprehensive comparison with other state-of-the-art methods in the field. A thorough literature review and comparison with existing approaches would enhance the paper's contribution and provide a clearer understanding of where the proposed method stands in the current landscape.

  While the paper concludes that the proposed system is promising for deployment in real smart home environments, there is a lack of explicit statements regarding the relevance of the research to practical applications. How well does the proposed method address real-world challenges and considerations, such as scalability, adaptability to diverse home environments, and potential integration issues? The paper should explicitly discuss these aspects to demonstrate its practical significance.

  Providing a more detailed roadmap for future research, including specific challenges to be addressed and potential innovations, would strengthen the paper's contribution and impact.

•  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The article "IoT Traffic Analysis and Device Identification Based on Two-stage Clustering in Smart Home Environments" contains information technical and innovative. The problem addressed is current and has technical relevance, which makes it significant.  

My recommendations are:

1. In the Introduction section and not only, the authors cite some works as a whole, e.g. [3-8], [9-11] but it would be good if each one were detailed. Also, they did not indicate the advantages or disadvantages of this work.

2. It will be better to give some explanation regarding the connection between the proposed method and the device. 

3. It would be good to score with the other related works the accuracy obtained, maybe some comparisons.

4. The references are old.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The paper "IoT Traffic Analysis and Device Identification Based on Two-stage Clustering in Smart Home Environments" can be accepted in current form.