EVFL: Towards Efficient Verifiable Federated Learning via Parameter Reuse and Adaptive Sparsification

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript entitled EVFL: Towards Efficient Verifiable Federated Learning via Parameters-Resuing and Adaptive Sparsification. The contribution of the article is the introduction of Efficient verifiable federated learning(EVFL), which integrates the methods of adaptive gradient sparsification (AdaGS), Boneh-Lynn-Shacham (BLS) signature, and fully homomorphic encryption (FHE).

   In the abstract, Federated model has to be explained clearly. The author has mentioned the cloud server responsible for aggregating model parameters may be malicious. Is that a public cloud or private cloud? On what basis this claim is made.  Do authors use any resources of cloud service providers ? Abstract is too lengthy and the important points are not clear.

 In the introduction and related works section, there are many research works on integrating CLOUD-FOG-EDGE COMPUTING model. That type of models is not referred.  I hope adding these points will add to the need of federated learning in IIOT. Moreover, the references are not well elaborated.

 In 2.1 System model,what are the potential threats, how the proposed methods will handle?elaborate

Refer to Figure 1. The FL system model. The picture icon has firewall symbol, is it used in the model ?

In System setup: What about the infrastructure set up for cloud server? Should elaborate the latency in transferring the data from cloud to the local node.

Section 6 : Data set , the source of the data should be elaborated with their parameters. (F-MNIST) and kdd cup.

Conclusion don’t have clear idea to the readers, should be re-written.

Some kind of words like Amalgamates, bolstering which do not fit well to this manuscript, however the meaning of the work is same.Consider writing appropriate words

Strength

1.       Good Idea to maintain security and integrity

2.       Good in proposed methods

3.       Good in theoretical approaches

 

Weakness

1.       Infrastructure set up for Federated learning can be improved

2.       No specific real time case studies in IIoT is mentioned

3.       Practical difficulties in implementation with real-world IIoT along with the proposed methods is missing

 

 

 

 

Comments on the Quality of English Language

Needs Improvement

Author Response

Please see the attachment for details.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

contribution:

Write the contribution made at the beginning of a new sentence followed by an explanation of the method used. Example for contribution 1: “we design an efficient verification mechanism by...”

Problem statement:

the content is not appropriate  :  At least contain these points: Identify the research problem that occurs in Federated Learning, Research Objectives, Specific questions to be answered through the research, Significance of the Research and Reasons why this research is important and how the results can contribute to knowledge

making work flow diagram in figure 4 does not use standardized references, for example process uses verbs and has different chart symbols compared to I/O.

The experimental results did not show the promised contribution. Develop an experiment that specifically addresses this:

1. efficient verification mechanism

2. schemes that are better suitable for resource-constrained environments.

3. effectiveness in protecting agent data privacy and achieves secure verification

Comments on the Quality of English Language

overall ok

Author Response

Please see the attachment for details.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Appreciate the authors effort to incorporate the comments.

 

Comments on the Quality of English Language

Still, a strong proof-read  is needed before final submission.