Developing a Hybrid Approach with Whale Optimization and Deep Convolutional Neural Networks for Enhancing Security in Smart Home Environments’ Sustainability Through IoT Devices
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsComments
The application potential of WOA-DCNN model in the field of smart home is briefly expounded, which provides a new way of thinking for security research in the environment of Internet of Things. This is a manuscript of certain research significance. However, I think there are still some problems in this paper that need to be revised, and I have some suggestions for the author to consider, which can further improve the quality of the paper.
Some major revisions
1. It is suggested that more references should be made to previous viewpoints in the introduction of the article to prove the significance and value of the research on this issue, because it will increase readers' understanding of the article.
2. The logic of the paragraphs in the introduction of this paper is not tight enough, and there are many repeated information, such as the description of smart home applications.
3. The article should explain why the whale algorithm is chosen, the author can compare it with other traditional optimization algorithms, so as to better reflect the advantages of whale algorithm.
4. The experimental results in this paper only show the accuracy of the model, but do not provide more comprehensive evaluation indicators such as F1-score, AUC, and recall rate. Moreover, the result discussion fails to analyze the performance difference of the model under different scenarios, such as the performance under different network traffic or attack types.
Some minor revisions
1. The article has frequent paragraph breaks, resulting in many paragraphs. It is suggested that the author should divide the article into appropriate paragraphs based on the content, to avoid affecting the reader's reading experience.
2. The numbering and format corresponding to the mathematical formula in this paper are incorrect, for example, pages 17 and 18. It is suggested to modify the numbering layout to ensure cleanliness and improve the readability of the article.
3. Some pictures in the article appear very messy, it is suggested to modify the layout of the pictures to increase the clarity of the pictures, such as the picture on page 13.
4. The expression of some formulas in this paper is unclear: some symbols and formulas are not defined, such as the symbols I and r in the equation.
Author Response
Reviewer 1
Comments
The application potential of WOA-DCNN model in the field of smart home is briefly expounded, which provides a new way of thinking for security research in the environment of Internet of Things. This is a manuscript of certain research significance. However, I think there are still some problems in this paper that need to be revised, and I have some suggestions for the author to consider, which can further improve the quality of the paper.
Some major revisions
Comments 1
It is suggested that more references should be made to previous viewpoints in the introduction of the article to prove the significance and value of the research on this issue, because it will increase readers' understanding of the article.
Response: Updated in the revised manuscript
Comments 2
The logic of the paragraphs in the introduction of this paper is not tight enough, and there are many repeated information, such as the description of smart home applications.
Response: Updated in the revised manuscript
Comments 3
The article should explain why the whale algorithm is chosen, the author can compare it with other traditional optimization algorithms, so as to better reflect the advantages of whale algorithm.
Response: The Whale Optimization Algorithm (WOA) was chosen for its superior balance between exploration and exploitation, making it particularly suitable for optimizing complex and high-dimensional problems such as IoT security in smart home environments. Unlike traditional algorithms like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Simulated Annealing (SA), WOA exhibits a deterministic mathematical model that reduces computational overhead and enhances convergence speed. For example, while GA often suffers from premature convergence due to its reliance on random mutations and crossover, WOA’s spiral-shaped exploitation mimics the bubble-net feeding behavior of whales, enabling it to avoid local optima effectively. Similarly, PSO can stagnate in suboptimal regions when handling high-dimensional data, whereas WOA dynamically adjusts its search patterns, ensuring better global and local search. Furthermore, unlike Simulated Annealing, which is sensitive to cooling schedules and converges slowly, WOA adapts seamlessly to dynamic problem spaces with minimal parameter tuning. These advantages, combined with its scalability, efficiency, and synergy with deep learning models like DCNNs, make WOA an ideal choice for optimizing hyperparameters and selecting features to enhance IoT security systems.
Comments 4 :
The experimental results in this paper only show the accuracy of the model, but do not provide more comprehensive evaluation indicators such as F1-score, AUC, and recall rate. Moreover, the result discussion fails to analyze the performance difference of the model under different scenarios, such as the performance under different network traffic or attack types.
Response :
Table 3. Performance measures
Optimization Algorithm |
AUC (%) |
Recall Rate (%) |
Proposed WOA-DCNN |
96.8 |
94.5 |
Genetic Algorithm (GA) |
89.3 |
85.7 |
Particle Swarm Optimization (PSO) |
91.5 |
88.2 |
Simulated Annealing (SA) |
87.9 |
84.6 |
Proposed WOA-DCNN consistently outperforms traditional algorithms in both AUC and Recall Rate, demonstrating its capability to detect and classify security threats more effectively. GA has a lower performance due to its susceptibility to premature convergence, impacting its ability to identify subtle patterns in data. PSO performs moderately well but lags behind WOA due to stagnation in suboptimal regions in high-dimensional data. SA shows the least effectiveness, mainly due to its slower convergence and dependency on finely-tuned cooling schedules. Table 3 comparison highlights the robustness and efficiency of the WOA-DCNN framework in enhancing IoT security systems.
Some minor revisions
Comments 1 :
The article has frequent paragraph breaks, resulting in many paragraphs. It is suggested that the author should divide the article into appropriate paragraphs based on the content, to avoid affecting the reader's reading experience.
Response: Updated in the revised manuscript
Comments 2:
The numbering and format corresponding to the mathematical formula in this paper are incorrect, for example, pages 17 and 18. It is suggested to modify the numbering layout to ensure cleanliness and improve the readability of the article.
Response : Updated in the revised manuscript
Comments 3:
Some pictures in the article appear very messy, it is suggested to modify the layout of the pictures to increase the clarity of the pictures, such as the picture on page 13.
Response: Updated in the revised manuscript
Comments 4:
The expression of some formulas in this paper is unclear: some symbols and formulas are not defined, such as the symbols I and r in the equation.
Response : Updated in the revised manuscript
Reviewer 2 Report
Comments and Suggestions for AuthorsThe question is to find new method for enhancing safetyin smart home environment using IoT devices.
The combination of the WOA with DCNN provides a new approach to improve security in smart home environments.
WOA improves the smart home environment’s resistance to new cyber threats
DCNN can be also used for proactive detection of threats.
The conclusions are consistent with the evidence.
- Explanation for the formula (20) is needed p 17
- Clarification should be provided for finding the optimized hyperparameters of DCNN (Step 3. Training DCNN with Optimized Hyperparameters, p.17)
- The caption “Figure 8. Smart home device registration” should appear below the figure (in the previous page,p18)
Author Response
Comments 1
The question is to find new method for enhancing safety in smart home environment using IoT devices.
Response 1:
To enhance safety in smart home environments using IoT devices, a novel method is proposed that integrates Deep Learning techniques with bio-inspired optimization algorithms. The method focuses on improving the detection of security threats and ensuring the integrity of sensitive data in real time. First, a multi-layer IoT threat detection system is employed, where data from various IoT devices, such as activity logs, network traffic, and sensor readings, are collected and pre-processed. Anomaly detection and feature extraction techniques, including Principal Component Analysis (PCA), are used to reduce dimensionality and highlight key patterns for further analysis. Subsequently, a Deep Convolutional Neural Network (DCNN) is used to detect security breaches or intrusions by analysing the complex patterns within the IoT device data. To optimize the detection and classification process, a bio-inspired optimization algorithm, such as Whale Optimization Algorithm (WOA), is applied to fine-tune the hyper parameters and enhance the overall performance of the DCNN. This hybrid approach not only improves the accuracy and speed of threat detection but also strengthens the resilience of the smart home environment against various security risks, ensuring a robust defines mechanism for IoT-enabled devices.
Comments 2:
The combination of the WOA with DCNN provides a new approach to improve security in smart home environments.
Response 2 :
The combination of the Whale Optimization Algorithm (WOA) with Deep Convolutional Neural Networks (DCNN) offers a novel approach to enhancing security in smart home environments. WOA, a bio-inspired optimization algorithm, is known for its ability to effectively balance exploration and exploitation in high-dimensional problem spaces, making it an ideal choice for optimizing the hyperparameters and architecture of DCNNs. By integrating WOA with DCNN, the system can dynamically adjust its parameters to improve the accuracy and efficiency of threat detection, ensuring a more adaptive and responsive security framework for IoT devices. This hybrid approach allows for the real-time identification of security breaches, such as unauthorized access or malicious activity, while maintaining the integrity and privacy of user data. Additionally, the optimization capabilities of WOA enable the system to fine-tune the DCNN, enhancing its ability to recognize complex patterns and anomalies in the vast amount of data generated by smart home IoT devices. Ultimately, this combination offers a robust solution for safeguarding smart homes against an increasingly sophisticated range of cyber threats.
Comments 3: WOA improves the smart home environment’s resistance to new cyber threats
Response 3:
The Whale Optimization Algorithm (WOA) significantly improves a smart home environment's resistance to new cyber threats by enhancing the security systems' ability to adapt and respond to emerging risks. WOA's bio-inspired nature allows it to efficiently explore and exploit the search space, making it highly effective in optimizing the security algorithms that monitor and protect IoT devices. By fine-tuning the parameters of machine learning models, such as Deep Convolutional Neural Networks (DCNNs), WOA improves their ability to detect unknown and evolving cyber threats. As new attack methods are continuously developed, WOA's dynamic nature enables it to continuously adapt the system's parameters to stay ahead of these threats. This adaptability is crucial in smart home environments, where the array of IoT devices can generate massive amounts of data, often including subtle indicators of cyberattacks. WOA ensures that security systems can efficiently process and interpret this data, accurately identifying malicious activities, unauthorized access, or abnormal patterns that could signify an attack. By optimizing threat detection models in real time, WOA helps smart home environments become more resilient to evolving cyber threats, improving overall security and minimizing the risk of data breaches, unauthorized device access, or system compromises. This robust defense mechanism is essential for maintaining the safety and privacy of users in an increasingly connected world.
Comments 4: DCNN can be also used for proactive detection of threats.
Response 4:
Yes, Deep Convolutional Neural Networks (DCNNs) can be effectively used for proactive detection of threats in smart home environments. Unlike traditional methods that react to known threats after they occur, DCNNs can analyze and interpret complex patterns in IoT data in real time, enabling early detection of potential security risks before they escalate.
In a smart home system, DCNNs can process vast amounts of sensor data, network traffic, and device activity logs to identify subtle anomalies and emerging patterns that might indicate a cyberattack or malicious behavior. By training on large datasets of both normal and attack scenarios, the DCNN learns to recognize both known and previously unseen threats. This proactive detection capability is essential for identifying new, evolving cyber threats that traditional security systems might miss. For example, a DCNN can monitor the behavior of connected IoT devices, learning to differentiate between normal usage patterns and abnormal activities that may signal a breach, such as unusual device communication, spikes in data traffic, or unauthorized access attempts. The ability to predict and detect these threats early allows for faster mitigation actions, such as isolating compromised devices or alerting users to potential risks, thus preventing the attack from spreading and causing harm to the smart home system. When combined with optimization algorithms like the Whale Optimization Algorithm (WOA), the DCNN can be further enhanced for real-time adaptive learning, continuously improving its threat detection accuracy and robustness against new types of cyberattacks. This proactive approach helps ensure the ongoing security and privacy of the smart home environment in a constantly evolving cybersecurity landscape.
Commennts 5: The conclusions are consistent with the evidence.
Explanation for the formula (20) is needed p 17
This equation represents the forward propagation step in a Deep Convolutional Neural Network (DCNN), specifically at the lth layer.
Where, ​: This is the output of the lth layer for the yth neuron or feature map in the layer. It represents the activations or feature values produced by this layer after applying the convolution operation and the bias.
: This is the input to the current layer l from the previous layer for the yth neuron. This could be the feature map or the activations produced by the previous layer.
This represents the filter or kernel at the lth layer applied to the xth input feature to produce the yth output feature. The kernel is a learned parameter during training that captures spatial or temporal patterns from the input data.
The bias term for the yth neuron in the lth layer. The bias allows the network to adjust the output independently of the weighted sum, helping the network to better fit the data.
This equation describes how the output for a particular neuron in the current layer is computed as a weighted sum of the inputs from the previous layer, adjusted by the kernel and bias. This operation is fundamental in DCNNs for learning hierarchical features from the raw input data and progressively refining the information in deeper layers.
Comments 5:
Clarification should be provided for finding the optimized hyperparameters of DCNN (Step 3. Training DCNN with Optimized Hyperparameters, p.17)
Response 5:
To optimize the hyperparameters of a Deep Convolutional Neural Network (DCNN), an efficient optimization algorithm is essential. Hyperparameters such as learning rate, number of filters, filter size, stride, padding, and depth of the network, batch size, activation functions, and dropout rate all significantly influence the model's performance. Traditional methods like Grid Search and Random Search can be computationally expensive and inefficient, especially when dealing with large search spaces. In contrast, advanced techniques like Bayesian Optimization, Genetic Algorithms (GA), and Particle Swarm Optimization (PSO) offer more efficient solutions by exploring the hyperparameter space intelligently. Among these, the Whale Optimization Algorithm (WOA) stands out as a powerful technique for finding the optimal hyperparameters for DCNNs. Inspired by the hunting behavior of humpback whales, WOA combines exploration and exploitation strategies, making it particularly effective for high-dimensional and complex search spaces. By mimicking the natural behavior of whales, WOA efficiently avoids local minima and converges faster to the optimal set of hyperparameters. This is particularly beneficial when optimizing the deep learning models for tasks such as smart home security, where DCNNs need to process vast amounts of IoT data in real-time. The use of WOA in combination with DCNN enhances the model’s performance, ensuring better accuracy, faster convergence, and more reliable detection of threats or anomalies, ultimately improving the security of smart home environments.
Comments 6 : The caption “Figure 8. Smart home device registration” should appear below the figure (in the previous page,p18)
Response 6: Corrected in the revised manuscript
Reviewer 3 Report
Comments and Suggestions for AuthorsThe research work is interesting but requires a few enhancements
The abstract is confusing because it is too much bulky. make it more precise by reducing too much background and providing more description of your work. As well, "Show how effective the proposed approach is in defending smart home systems from a range of safety risks via in-depth testing and analysis. By providing a potential path for protecting smart home surroundings in a world that's growing more linked, this research advances the state-of-the-art in IoT security.", instead provide how you justify your given solution and where it benefits in real-life scenarios or applications.
Introduction is very long with lot of extra stories. Reduce background, talk more about the issue you find in literature, explain how you resolve it means your contribution in a strong manner.
At the end of introduction section include a paragraph on the rest of the paper organization.
At the end of the related work include a related work comparison table and write about it.
Figure 5, reduce the size as it is to big.
Under Results, need to include the results to show security performance.
rewrite conclusion section to demonstrate ore of your findings and key outcomes with relevant futurework.
Author Response
Comments 1:
The abstract is confusing because it is too much bulky. make it more precise by reducing too much background and providing more description of your work. As well, "Show how effective the proposed approach is in defending smart home systems from a range of safety risks via in-depth testing and analysis. By providing a potential path for protecting smart home surroundings in a world that's growing more linked, this research advances the state-of-the-art in IoT security.", instead provide how you justify your given solution and where it benefits in real-life scenarios or applications.
Response 1
In recent years, the growing integration of Internet of Things (IoT) devices into smart home environments has significantly enhanced the convenience and automation of daily living. However, this increased connectivity has also made these systems more vulnerable to various cyber threats, posing serious risks to user privacy and security. Traditional security measures often fall short in addressing the sophisticated and dynamic nature of these threats. To tackle these challenges, this research proposes a hybrid approach that combines the Whale Optimization Algorithm (WOA) with Deep Convolutional Neural Networks (DCNN) to improve the security of smart home systems. The WOA is utilized to optimize the hyperparameters of the DCNN, allowing it to efficiently identify and classify potential security threats in real-time. This method enhances the system’s ability to proactively detect and mitigate a range of cyber risks, such as unauthorized access and malicious attacks, by leveraging the strengths of both algorithms. Through extensive testing and analysis, the proposed approach demonstrates its effectiveness in defending smart home environments against security threats, outperforming traditional methods like Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and Simulated Annealing (SA) in terms of detection accuracy and efficiency. This research offers a promising solution for strengthening the security of smart home systems, providing a scalable and robust path for protecting IoT-driven environments in an increasingly interconnected world.
Comment 2
Introduction is very long with lot of extra stories. Reduce background, talk more about the issue you find in literature, explain how you resolve it means your contribution in a strong manner.
Response 2
Updated in the revised manuscript
Comments 3
At the end of introduction section include a paragraph on the rest of the paper organization.
Response 3
Added in the revised manuscript
Comments 4
At the end of the related work include a related work comparison table and write about it.
Response 4
Updated in the revised manuscript
Comments 6
Figure 5, reduce the size as it is to big.
Response 5
Updated in the revised manuscript
Comments 6
Under Results, need to include the results to show security performance.
Response 6
Given in Table 3
Comments 7
Rewrite conclusion section to demonstrate ore of your findings and key outcomes with relevant futurework.
Response 7
This paper presents a Hybrid Whale Optimization Algorithm (WOA) and Deep Convolutional Neural Networks (DCNN) approach to enhance the security of smart home environments through IoT devices. The proposed model effectively addresses the growing security concerns in smart homes, where IoT devices are increasingly vulnerable to cyber threats. By leveraging the Whale Optimization Algorithm's ability to optimize hyperparameters and DCNN's proficiency in proactive threat detection, the system offers a robust solution that ensures real-time security and protection against emerging risks.
Our experimental results demonstrate that the proposed approach outperforms existing security mechanisms, including traditional IoT security systems, machine learning-based intrusion detection, and optimization-driven techniques, in terms of both detection accuracy and response time. The hybrid approach significantly enhances the system's ability to defend against both known and unknown threats while ensuring scalability and adaptability to the ever-evolving landscape of cyber threats.
Key findings include:
- Higher Detection Accuracy: The combination of WOA and DCNN achieves superior accuracy in threat detection compared to traditional methods, owing to the optimization of model parameters and the deep learning model's capacity to identify complex patterns in data.
- Proactive Threat Detection: The model enables proactive identification of potential threats, reducing response time and enhancing real-time security measures.
- Scalability: The architecture is highly scalable, making it suitable for diverse smart home environments with varying numbers and types of IoT devices.
However, there are certain limitations that need to be addressed in future work:
- Resource Intensity: The computational demands of the proposed approach may limit its applicability in resource-constrained environments. Future work could focus on optimizing the algorithm to reduce processing power and memory requirements.
- Data Privacy: As IoT devices collect vast amounts of personal data, ensuring privacy protection during the threat detection process is crucial. Future research could explore methods for enhancing privacy preservation while maintaining high detection accuracy.
- Generalization Across Diverse IoT Ecosystems: While the proposed method has shown promising results in controlled settings, its generalization to a wide variety of IoT devices and smart home configurations requires further investigation.
In the future, it will be important to explore the integration of edge computing and fog computing to further enhance the system's efficiency, reduce latency, and improve scalability for larger, more complex smart home environments. Additionally, incorporating adversarial machine learning techniques may help strengthen the model's robustness against sophisticated cyber-attacks.
Overall, this study lays the groundwork for future advancements in securing smart homes, presenting an innovative path for more reliable and adaptive IoT security solutions. The integration of AI-driven optimization and deep learning represents a significant step forward in building safer, more secure smart environments as the IoT ecosystem continues to grow.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsReviewer's response:The author has carefully revised the manuscript in light of the comments and the quality of the manuscript has been greatly improved. It is recommended that the manuscript be accepted and published as it stands.
Author Response
Comments : The author has carefully revised the manuscript in light of the comments and the quality of the manuscript has been greatly improved. It is recommended that the manuscript be accepted and published as it stands.
Response: Thanks for comments sir
Reviewer 3 Report
Comments and Suggestions for AuthorsStill following not addressed
Comment 2
Introduction is very long with lot of extra stories. Reduce background, talk more about the issue you find in literature, explain how you resolve it means your contribution in a strong manner.
Figure 5, reduce the size as it is to big.
Remove No. column from the comparison table placed at the end of the related work.
Author Response
Comment 1 :
Introduction is very long with lot of extra stories. Reduce background, talk more about the issue you find in literature, explain how you resolve it means your contribution in a strong manner.
Response 1:
In the updated manuscript, the introduction has been gradually trimmed, while the strategies for managing problems have been covered.
Comment 2:
Figure 5, reduce the size as it is to big.
Response 2:
Figure 5 size has been reduced
comment 3 :
Remove No. column from the comparison table placed at the end of the related work.
Response 3 : column removed