Cloud-Based Smart Contract Analysis in FinTech Using IoT-Integrated Federated Learning in Intrusion Detection
Round 1
Reviewer 1 Report
Dear Author(s),
Please find below my concerns and recommendations regarding your manuscript proposal entitled "Cloud based smart contract analysis in fintech using IoT integrated federated learning in intrusion detection" sent to Data MDPI Journal.
First of all, you should know that before writing this review, I conducted a self documentation and I found that small parts of your texts are similar to some articles already published:
- the sequences between the rows 134 - 139, 141 - 145 are quite similar to the article available at the adress: https://doi.org/10.3390/electronics9010094
- the sequence between the rows 181 - 187 are quite similar to the manuscript available at the address: https://doi.org/10.48550/arXiv.2007.09712
- the sequence between the rows 287 - 290 are similar to the article published at: https://doi.org/10.1007/978-3-319-52727-7_11
- the sequence between the rows 397 - 408 are quite similar to the article available at: https://doi.org/10.1155/2021/9361348.
I recommend you to revise and rewrite these sequences so that you decrease the level of similarity.
Within the abstract, at the row 25, you use the acronym "AUC". I recommend you to describe each acronym at its first appearance.
The same remark for the acronym "ML" at the row 31.
At the rows 37 - 38 you say: "The Overall Information Insurance Guideline (GDPR) is among the moves made because of information security across globe." Please pay attention, because GDPR comes from "General Data Protection Regulation", not from "Overall Information Insurance Guideline". Please revise and correct it.
The Introduction chapter should be seriously improved. Here you should clearly define and describe the following important aspects:
- the research gap;
- the research goal;
- the research question(s).
The readers are interested to know from the very beginning of the manuscript what you want to cover by your research proposal.
At the rows 103 - 105 you say: "In this section, a novel method for cloud-based smart contracts in Fintech data and their intrusion detection with a cyber threat federated graphical authentication system are proposed and discussed." I recommend you to present some arguments for the "novelty" of your method.
I recommend you to include in your work the following relevant resources: https://doi.org/10.3390/sym11030340, https://doi.org/10.15240/tul/001/2021-2-007, https://doi.org/10.3390/su11051419, https://doi.org/10.3390/app112110353. By including these references in your article, you will widen the general context of your research proposal.
In the Conclusions chapter, please present the following aspects:
- the managerial implications (here is the place where you can "sell" your research results to the readers);
- the limitations of your research proposal.
Dear Author(s),
Please consider all the above remarks as being constructive recommendations in order to improve the general quality of your manuscript proposal.
Kind Regards!
Author Response
Please find below my concerns and recommendations regarding your manuscript proposal entitled "Cloud based smart contract analysis in fintech using IoT integrated federated learning in intrusion detection" sent to Data MDPI Journal.
First of all, you should know that before writing this review, I conducted a self documentation and I found that small parts of your texts are similar to some articles already published:
- the sequences between the rows 134 - 139, 141 - 145 are quite similar to the article available at the adress: https://doi.org/10.3390/electronics9010094
- the sequence between the rows 181 - 187 are quite similar to the manuscript available at the address: https://doi.org/10.48550/arXiv.2007.09712
- the sequence between the rows 287 - 290 are similar to the article published at: https://doi.org/10.1007/978-3-319-52727-7_11
- the sequence between the rows 397 - 408 are quite similar to the article available at: https://doi.org/10.1155/2021/9361348.
I recommend you to revise and rewrite these sequences so that you decrease the level of similarity.
Response: As per above comments the article has been enhanced by decreasing the similarity.
Within the abstract, at the row 25, you use the acronym "AUC". I recommend you to describe each acronym at its first appearance.
Response: acronym for AUC is added and highlighted as “AUC (Area Under Curve”
The same remark for the acronym "ML" at the row 31
Response: The acronym is added for ML and highlighted as “ML (machine Learning)”
At the rows 37 - 38 you say: "The Overall Information Insurance Guideline (GDPR) is among the moves made because of information security across globe." Please pay attention, because GDPR comes from "General Data Protection Regulation", not from "Overall Information Insurance Guideline". Please revise and correct it.
Response: as per above comments GDPR is corrected as General Data Protection Regulation and highlighted.
The Introduction chapter should be seriously improved. Here you should clearly define and describe the following important aspects:
Response:
Research gap:
The research gap is added as “Ongoing advancements in deep learning procedures have delivered critical enhancements in well established simulated intelligence occupations like medication revelation, quality examination, and discourse and picture acknowledgment. According to McMahan et al., despite the numerous benefits of deep learning, the same training dataset that made it so reliable also raises serious privacy concerns. developed Federated Deep Learning, a mobile device-specific distributed deep learning paradigm (FDL). In FDL, distributed training involves a number of parties, and a parameter server keeps track of a developing deep learning model. This effectively combines distributed computation with deep learning.”
Research goal:
The research goal is added as “the proposed model is developed for cloud-based smart contracts in Fintech data and their intrusion detection with a cyber threat federated graphical authentication system. Numerous advantages result from the use of blockchain technology and artificial intelligence. In brief, by using machine learning's analytical capabilities, the built-in security features of blockchain can be reinforced. The ability to process massive amounts of data securely and effectively in the financial services industry may be extremely valuable to institutions and end users.”
Research question:
Due to the rapid development of wireless transmission and processing, the industrial Internet of Things (IIoT) has advanced significantly over the past few years. On IoT networks, a variety of cutting-edge portable devices, including smart phones, smart watches, and smart applications, have emerged. These have been extensively utilized by numerous businesses, including live gaming, smart manufacturing, navigational systems, smart cities, and smart healthcare. Due to their rapid proliferation, the architecture of IoT networks still faces a number of significant obstacles. One of the main obstacles is coming up with effective and adaptable control for IoT systems that can help save energy, increase the number of applications, and make it easier to expand in the future. The IoT networks are cognitively demanding, time-efficient, and constantly require computing resources, which is another significant barrier. They also guarantee privacy and security against unauthorized access. People are beginning to consider personal data security even more seriously as a result of the rapid growth of personal awareness and digital technology. In order for devices to work together to create a single method of learning with local training, distributed learning methods are required. A decentralized platform for machine learning (ML) is federated learning (FL). Because data created on an end device does not leave the device, unlike centralized learning frameworks, the FL framework automatically promotes
distributed learning model on the device itself. A client device (e.g., a neighbourhood Wi-Fi switch) and a cloud server just offer the settings that have been changed. A portion of the advantages of involving FL in remote IoT networks are: ( i) local ML system settings can save power and use less wireless bandwidth than exchanging a lot of training data; ii) a significant reduction in transmission delay can be achieved by locally calibrating an ML model's parameters; ( iii) FL can assist with safeguarding information protection in light of the fact that main the neighbourhood learning model factors are sent and the preparation information stays on the edge devices.
The readers are interested to know from the very beginning of the manuscript what you want to cover by your research proposal.
Response: the proposed model for the research is based on cloud-based smart contracts in Fintech data and their intrusion detection with a cyber threat federated graphical authentication system.
At the rows 103 - 105 you say: "In this section, a novel method for cloud-based smart contracts in Fintech data and their intrusion detection with a cyber threat federated graphical authentication system are proposed and discussed." I recommend you to present some arguments for the "novelty" of your method.
Response: the novelty of this research is to develop the proposed model for cloud-based smart contracts in Fintech data and their intrusion detection
I recommend you to include in your work the following relevant resources: https://doi.org/10.3390/sym11030340, https://doi.org/10.15240/tul/001/2021-2-007, https://doi.org/10.3390/su11051419, https://doi.org/10.3390/app112110353. By including these references in your article, you will widen the general context of your research proposal.
Response: the above references has been added and highlighted
In the Conclusions chapter, please present the following aspects:
- the managerial implications (here is the place where you can "sell" your research results to the readers);
- the limitations of your research proposal.
Response: the conclusion presents a detailed discussion of Blockchain and AI technologies that will help form a sustainable smart society. We discuss blockchain security enhancement solutions, summarizing the key points that can be used for developing various blockchain-AI based intelligent transportation systems.
The limitation is added in conclusion and highlighted.
Reviewer 2 Report
The authors proposed a cloud-based IDS based on IoT on top of federated learning and smart contract. The idea is interesting. However, I have the following concerns.
1. Please revise the grammatical issues of the paper.
2. Recent FL-based papers are missing. Some are mentioned as follows.
-> "Federated Semisupervised Learning for Attack Detection in Industrial Internet of Things," in IEEE Transactions on Industrial Informatics, vol. 19, no. 1, pp. 286-295, Jan. 2023, doi: 10.1109/TII.2022.3156642.
-> "FBI: A Federated Learning-Based Blockchain-Embedded Data Accumulation Scheme Using Drones for Internet of Things," in IEEE Wireless Communications Letters, vol. 11, no. 5, pp. 972-976, May 2022, doi: 10.1109/LWC.2022.3151873.
3. Contribution of this paper is limited. FL-based intrusion detection is a very popular topic and well-explored. Please highlight your novelty based on the limitation of existing works.
4. Split the introduction into separate paragraphs for better readability. Moreover, follow the hourglass formula to revise the introduction.
5. Add a summary table for related works (i.e., Section 2) including the limitation of each.
6. Instead of making it general, authors specifically choose Ethereum. Please justify it.
7. Author considered a cloud-based scheme. How about a single point of failure or latency or higher energy consumption?
8. Please improve the quality of Fig 3.
9. A discussion on convergence is required.
10. Reference is required for equations.
11. Please add an algorithm section for better understanding.
12. A computational complexity analysis is required.
13. A computational complexity analysis is required.
14. Threat model is missing. A security analysis is required based on threats.
15. How many clients were considered? what are their computational resources?
16. Authors didn't consider the privacy of the model itself.
17. A Comparison with existing works is missing.
18. A summary table for simulation parameters is required.
19. Which consensus did the author consider? Performance metrics regarding blockchain are missing.
20. Which deep learning did they consider in the training is missing?
Author Response
The authors proposed a cloud-based IDS based on IoT on top of federated learning and smart contract. The idea is interesting. However, I have the following concerns.
- Please revise the grammatical issues of the paper.
Response: The grammar is enhanced as per above comments
- Recent FL-based papers are missing. Some are mentioned as follows.
-> "Federated Semisupervised Learning for Attack Detection in Industrial Internet of Things," in IEEE Transactions on Industrial Informatics, vol. 19, no. 1, pp. 286-295, Jan. 2023, doi: 10.1109/TII.2022.3156642.
-> "FBI: A Federated Learning-Based Blockchain-Embedded Data Accumulation Scheme Using Drones for Internet of Things," in IEEE Wireless Communications Letters, vol. 11, no. 5, pp. 972-976, May 2022, doi: 10.1109/LWC.2022.3151873.
Response: The above reference has been added and highlighted.
- Contribution of this paper is limited. FL-based intrusion detection is a very popular topic and well-explored. Please highlight your novelty based on the limitation of existing works.
Response: the limitation based on proposed technique is added in conclusion and highlighted as It is basic to plan correspondence productive methodologies that convey short messages or model changes over and over as a component of the preparation interaction, as opposed to sending the total informational collection over the organization. Connection, bandwidth, and power are essential for maintaining these activities because this process is carried out in millions of tiny devices. The best two choices for making these cycles more effective and diminishing correspondence in stages are: 1) reducing the total number of rounds of communication; and 2) reducing the number of messages exchanged during each round.
- Split the introduction into separate paragraphs for better readability. Moreover, follow the hourglass formula to revise the introduction.
Response: The introduction has been updated as per above comments
- Add a summary table for related works (i.e., Section 2) including the limitation of each. As
Response: as per above comments the summary table is added and highlighted as Table-1 Summarization based on existing FL based fintech intrusion detection
- Instead of making it general, authors specifically choose Ethereum. Please justify it.
Response: section 3.1 shows Ethereum based justification is discussed.
- Author considered a cloud-based scheme. How about a single point of failure or latency or higher energy consumption?
Response: the cloud scheme is given in section 3.1and the parameters discussed are accuracy, precision, RMSE, recall, F-measure, AUC
- Please improve the quality of Fig 3.
Response: The figure quality is enhanced as per above comments
- A discussion on convergence is required.
Response: The discussion is added and highlighted as Five edge nodes, fifteen end devices, and a parameter server make up the system in the experiment. Each of the five resource blocks contains an edge node, a device that is close by, a device that is relayed, and a device that is far away. Because system is set up for full participation, all devices, regardless of whether they have more or less data, must participate in model training. For a single resource block, the SNR is 30, the path loss factor is 2, and the channel bandwidth is set to W = 20. Default values for data quantity and data quality in simulation experiments are Q = 512 and = 0.5, respectively. On Windows 10, we use PyTorch 1.8 to carry out our experiments. We utilized a fully connected multi-layer processor with two hidden 64-unit output layers and headers to represent the policy. The actor network server's pricing strategy is determined by three headers. There are three headers for near and far devices and two headers for relay devices in the devices actor network to determine their communication strategies. Value coefficient is c = 0.5, the clip ratio is equal to 0.2, the discount factor is equal to 0.99, and local learning rate is set to 0.0003. Number of training episodes is set to E = 100, each episode has a length of K = 16, and number of previous experiences is L = 4. Depending on its computation resource, each device can select a different Db 16, 32 of minibatch size when performing the local update. The neural networks are then optimized using the Adam optimizer.
- Reference is required for equations.
Response: The references have been added for equation in 11, 12, 13 and highlighted
- Please add an algorithm section for better understanding.
Response: As per above comments the algorithm is added
- A computational complexity analysis is required.
Response: since FL is the main focus of this research computational analysis obtains very minimal value in comparison with existing technique.
- A computational complexity analysis is required.
Response: since FL is the main focus of this research computational analysis obtains very minimal value in comparison with existing technique.
- Threat model is missing. A security analysis is required based on threats.
Response: general security analysis is carried out and it is not specific security model
- How many clients were considered? what are their computational resources?
Response: the number of clients has been taken based on number of iterations in proposed technique
- Authors didn't consider the privacy of the model itself.
Response: by developing security model, the privacy is also enhanced
- A Comparison with existing works is missing.
Response: table 3 shows the comparative analysis with existing technique
- A summary table for simulation parameters is required.
Response: the simulation setup is discussed in section 4 and it is highlighted
- Which consensus did the author consider? Performance metrics regarding blockchain are missing.
Response: the performance has been analysed based on intrusion dataset for accuracy, precision, RMSE, recall, F-measure, AUC which gives the attack analysis using proposed blockchain model
- Which deep learning did they consider in the training is missing?
Response: FL is part of deep learning which is described in proposed model
Round 2
Reviewer 1 Report
Dear Author(s),
I have read the revised version of your manuscript proposal, but I found that you didn't address all my constructive recommendations from the previous round of review.
Please find below my concerns and I recommend you to address them one by one.
1. There are some visible similarities between sequences from your text and some published articles. For example:
- the rows 86 - 109 are very similar to the material available at the address: https://doi.org/10.3390/network3010008
- the rows 406 - 421 are similar to https://doi.org/10.1109/JIOT.2021.3081626.
- the rows 463 - 474 are compiled from https://doi.org/10.1155/2021/9361348 and https://doi.org/10.3390/sym11040583.
Please revise and correct this issues.
2. In the previous review report, I recommended you to include in your work the following useful resources: https://doi.org/10.3390/sym11030340, https://doi.org/10.15240/tul/001/2021-2-007, https://doi.org/10.3390/app112110353. Please include them in the revised version of your manuscript.
3. There is a typo-error in the title at the row 182: "3.1. Cloud based smart contarct analysis". "Contract" instead of "contarct". Please revise.
4. In the Discussion chapter you should present a comparison between your research results and the others' results from the scientific literature. This way, you will highlight your own contribution to the field of knowledge.
5. The Conclusion section should be improved by including the following aspects:
- the managerial implications of your research results;
- the research limitations;
- the future research directions.
Kind Regards!
Author Response
- There are some visible similarities between sequences from your text and some published articles. For example:
- the rows 86 - 109 are very similar to the material available at the address: https://doi.org/10.3390/network3010008
- the rows 406 - 421 are similar to https://doi.org/10.1109/JIOT.2021.3081626.
- the rows 463 - 474 are compiled from https://doi.org/10.1155/2021/9361348 and https://doi.org/10.3390/sym11040583.
Please revise and correct this issues.
Response: As per above comments the contents has been revised and updated
- In the previous review report, I recommended you to include in your work the following useful resources: https://doi.org/10.3390/sym11030340, https://doi.org/10.15240/tul/001/2021-2-007, https://doi.org/10.3390/app112110353. Please include them in the revised version of your manuscript.
Response: As per above comments the references has been added in reference-2,3,4
- There is a typo-error in the title at the row 182: "3.1. Cloud based smart contarct analysis". "Contract" instead of "contarct". Please revise.
Response: The typo error has been corrected in the section 3.1 as per above comments
- In the Discussion chapter you should present a comparison between your research results and the others' results from the scientific literature. This way, you will highlight your own contribution to the field of knowledge.
Response: The results have been added in discussion part as per above comments as per the proposed contribution
- The Conclusion section should be improved by including the following aspects:
- the managerial implications of your research results;
- the research limitations;
- the future research directions.
Response: The results have been added for proposed technique. The limitation has been added in conclusion. The future scope is added and highlighted
Reviewer 2 Report
Thank for you revising the paper. Still, I have the following concerns.
- Which consensus did the author consider? Performance metrics regarding blockchain are missing.
Response: the performance has been analysed based on intrusion dataset for accuracy, precision, RMSE, recall, F-measure, AUC which gives the attack analysis using proposed blockchain model.
Comment: They are related to ML/DL not related to blockchain.
- Which deep learning did they consider in the training is missing?
Response: FL is part of deep learning which is described in proposed model.
Comment: FL is a framework on which DL model trains privately. Which DL model did you consider during the training?
Author Response
- Which consensus did the author consider? Performance metrics regarding blockchain are missing.
Response: the performance has been analysed based on intrusion dataset for accuracy, precision, RMSE, recall, F-measure, AUC which gives the attack analysis using proposed blockchain model.
Comment: They are related to ML/DL not related to blockchain.
Response: The parameters based on blockchain is explained as per above comments. The parameter for blockchain are trust value, scalability, integrity.
- Which deep learning did they consider in the training is missing?
Response: FL is part of deep learning which is described in proposed model.
Comment: FL is a framework on which DL model trains privately. Which DL model did you consider during the training?
Response: The training is carried out using FL and its network training is added and highlighted.
Round 3
Reviewer 1 Report
Dear Author(s),
For the revised version of your article, I have the following minor recommendations:
- some sentences starts with small letter. Please update the text, so that all the sentences start with capital letters. See, for example: row 618 ("proposed technique..."), row 26.
- there are some extra spaces between rows 286 and 287, 488 and 489. Please revise them.
- in the Conclusion chapter, please also present your research limitations.
Kind Regards!
Author Response
For the revised version of your article, I have the following minor recommendations:
- some sentences start with small letter. Please update the text, so that all the sentences start with capital letters. See, for example: row 618 ("proposed technique..."), row 26.
Response: The sentences has been updated as per above comments
- there are some extra spaces between rows 286 and 287, 488 and 489. Please revise them.
Response: The above comments has been updated
- in the Conclusion chapter, please also present your research limitations.
Response: The limitation is added in the conclusion part: “It is basic to plan correspondence productive methodologies that convey short messages or model changes over and over as a component of the preparation interaction, as opposed to sending the total informational collection over the organization. Connection, bandwidth, and power are essential for maintaining these activities because this process is carried out in millions of tiny devices. The best two choices for making these cycles more effective and diminishing correspondence in stages are: 1) reducing the total number of rounds of communication; and 2) reducing the number of messages exchanged during each rou
