Next Article in Journal
AI-MDD-UX: Revolutionizing E-Commerce User Experience with Generative AI and Model-Driven Development
Next Article in Special Issue
An Explainable Machine Learning Approach for IoT-Supported Shaft Power Estimation and Performance Analysis for Marine Vessels
Previous Article in Journal
Detecting Disinformation in Croatian Social Media Comments
Previous Article in Special Issue
Optimizing IoT Video Data: Dimensionality Reduction for Efficient Deep Learning on Edge Computing
 
 
Article
Peer-Review Record

Edge AI for Real-Time Anomaly Detection in Smart Homes

Future Internet 2025, 17(4), 179; https://doi.org/10.3390/fi17040179
by Manuel J. C. S. Reis 1,2,* and Carlos Serôdio 1,3
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Future Internet 2025, 17(4), 179; https://doi.org/10.3390/fi17040179
Submission received: 22 March 2025 / Revised: 14 April 2025 / Accepted: 16 April 2025 / Published: 18 April 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors


This manuscript presents an Edge AI-based framework for anomaly detection in smart home environments using a hybrid approach that combines Isolation Forest (IF) and Long Short-Term Memory Autoencoder (LSTM-AE). The study is timely, addresses a relevant research problem, and makes a clear case for the benefits of real-time, privacy-preserving, and energy-efficient anomaly detection at the edge. However, the manuscript has several critical issues that must be addressed before publication. These include unresolved editorial placeholders, reliance on synthetic data without real-world validation, lack of mathematical rigor, and inconsistencies in reporting quantitative results.


1. Incomplete Quantitative Reporting  
   The LSTM-AE inference time is described as “X milliseconds” (Line 350), indicating missing values. Such placeholders need to be replaced with actual measurements to support the study’s conclusions.

2. Lack of Mathematical Formulation  
   Despite the technical nature of the work, the paper does not include mathematical definitions of key models (e.g., anomaly scoring mechanism for IF, reconstruction loss for LSTM-AE). At least a high-level equation would improve the scientific rigor.

3. Synthetic-Only Experimentation  
   The experiment relies entirely on simulated data. While it provides a good sandbox to validate ideas, generalizability is limited. Inclusion of experiments using public datasets (e.g., CASAS, BoT-IoT) is highly encouraged for robustness.

4. Minor Issues and Suggestions
- Threshold Selection (Line 536): Using the 97th percentile to set the anomaly threshold is arbitrary. Consider including a sensitivity analysis.
- Code Sections: While the code snippets are clear, they may be too detailed for the main body of the paper. Consider moving them to an appendix or supplementary material.
- Energy Metrics: The power savings claims using INA219 are interesting but lack details like sampling intervals or variability.
- Repetitive Language: Phrases like “real-time anomaly detection” and “privacy-preserving” are repeated excessively—consider editing for conciseness.

Author Response

Please refer to the attached report.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Overall, the paper is good. There are only minor issues that should be resolved in my view:

-The relationship of this paper to the start of the art should be described better. With just one reference in that section, I am not convinced of the novelty of the paper.

-The objectives/aims should be re-visited in the conclusion 

-There should be some discussion on if this paper is applicable for only Smart Homes - or perhaps Smart Buildings more generally.

Author Response

Please refer to the attached report.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author has diligently made improvements to the paper based on the reviewer's questions and requests. 

Back to TopTop