Privacy Preserving Human Mobility Generation Using Grid-Based Data and Graph Autoencoders

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper addresses a very important issue. In particular, it proposes a new technique for the generation of a synthetic set of trajectories starting from a reduced collection of data regarding personal mobility tracking, collected in the metropolitan area of Munich in Germany.

This problem is very important since the issue of privacy protection of mobility data limits the use of such datasets in many applications.

 

Regardless of this positive aspect, the paper presents some significant weak points:

- the proposed method is not compared to previously proposed techniques and the experiments only focus on the evaluation of the properties of its application to the Munich case study.

- there is no consideration of the applicability of the proposed methods to other cases.

- Section 3, and in particular subsections 3.3, 3.4, and 3.5, are too informal. A more formal presentation of the contribution could radically increase its readability and clarity.

- writing quality of English must be improved

 

More detailed comments:

- line 57: You cite both figure 1 and figure 3, but here, the reader can better understand figure 1, figure 3 is too complicated. Moreover, in Figure 1 the modules “Metrices for Single Data Points” and “Aggregated Dataset Metrices” are present, but they are not described in the text at all. You should add at least one sentence to explain the content of Figure 1.

- line 198: hat -> that

- line 219: section ??, please report the section number correctly

- section 3.3: here some formal definition of the solution should be presented. How is a trajectory represented? How is the graph structure used for the representation of the trajectory? …

- section 3.4: In this section, it is necessary to show the architecture of the model (neural network), before explaining the latent space creation.

- line 234: Figure 3.3 does not exist! Is it figure 4?

- line 254: chapter -> section

. line 282: It is unclear where this loss function is applied, to which part of the model, and how. You cannot show equations here, at page 9 of your paper, without any formal definition of the neural network. Moreover, you did not put labels that allow you to cite the equation in the text; see line 285.

- line 285: The labels (2) and (3) are not present in the paper.

- line 299-320: This pseudo-code presentation is useless if you do not provide a formal representation of your data and of your model architecture.

- section 3.6: Also in this section a more formal presentation would be more effective.

- Figure 8: the cluster view is on the right!!! And the synthetically generated tracking data on the left side!!!

Comments on the Quality of English Language

English must be improved throughout all sections of the paper. Sentences are often too long, and there is no subdivision in paragraphs.

Author Response

Dear Reviewer;
Thank you very much for your time and your valuable feedback our article “Privacy Preserving Human Mobility Generation using Grid based Data and Graph Autoencoders”, submitted to the ISPRS International Journal of Geo-Information.
Please find our responses in the attached letter. We hope that we have been able to answer all points to your satisfaction and we look forward to receiving your positive feedback for publication.

Yours sincerely,
Fabian Netzler (corresponding author) on behalf of the authors

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Comments for author File: Comments.pdf

Author Response

Dear Reviewer;

Thank you very much for your time and your valuable feedback our article “Privacy Preserving Human 
Mobility Generation using Grid based Data and Graph Autoencoders”, submitted to the ISPRS International Journal of Geo-Information. Please find our responses in the attached letter.
We hope that we have been able to answer all points to your satisfaction and we look forward to receiving your positive feedback for publication.

Yours sincerely,
Fabian Netzler (corresponding author) on behalf of the authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper presents an algorithm for privacy preserving mobility generation. 

The paper is overall well structured but there are parts that need to be improved.

1. I do not undestand how the described privacy distance works. I think is is important to better describe how do you measure privacy and how does your algorthm comply with it. The simple distance between points as in figure 9 and 10 seems a rather weak definition. I would like to understand why the procedure cannot be inverted to retrieve original lcations

2. There are many works dealiing with trjectories generation. I think that a comparison and more punctual analysis on how your appraoch improves the state of the art would be important.

 

 

Comments on the Quality of English Language

I think that some sentences can be simplified: e.g.,

"The plot giving the distance between the original and synthetic data shows how far away an end or starting point of an original way is compared to the same way in the artificial dataset. A point is only compared to its counterpart, so it is still possible that the trajectories are nearer without breaking the privacy requirements, for example, if the working place in the synthetic data set is near the home area in the original data"

 

 

Author Response

Dear Reviewer;

Thank you very much for your time and your valuable feedback our article “Privacy Preserving Human Mobility Generation using Grid based Data and Graph Autoencoders”, submitted to the ISPRS International Journal of Geo-Information.
Please find our responses in the attached letter. We hope that we have been able to answer all points to your satisfaction and we look forward to receiving your positive feedback for publication.

Yours sincerely,
Fabian Netzler (corresponding author) on behalf of the authors

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

1. It is suggested that experiments be carried out with other datasets from other cities with different road network topology in order to identify aspects of to identify areas for improvement of the method

2. It is suggested to address the impact on the proposed method of:

a. The frequency of collection of GPS data that form the trajectory, and 

b. The treatment of the noise (e.g. kalman filter) that could be generated from the trajectory.

3. It would be important to analyse how the proposed method would work in the presence of other variables such as congested roads, busy traffic times, traffic jams, weekends, etc.

4. On the basis of the results, it is suggested to propose future work to be carried out to address other aspects that have an impact on the method. 

Comments on the Quality of English Language

minor corrections

Author Response

Dear Reviewer;

Thank you very much for your time and your valuable feedback our article “Privacy Preserving Human Mobility Generation using Grid based Data and Graph Autoencoders”, submitted to the ISPRS International Journal of Geo-Information.
Please find our responses in the attached letter. We hope that we have been able to answer all points to your satisfaction and we look forward to receiving your positive feedback for publication.

Yours sincerely,
Fabian Netzler (corresponding author) on behalf of the authors

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Authors have improved the paper significally, I consider the paper is suitable to be publish