@misc{berke2022generating,
title = {Generating synthetic mobility data for a realistic population with {RNNs} to improve utility and privacy},
author = {Berke, Alex and Doorley, Ronan and Larson, Kent and Moro, Esteban},
year = {2022},
url = {http://arxiv.org/abs/2201.01139},
doi = {10.1145/3477314.3507230}
}

DL components:
RNN, LSTM

Evaluation metrics:
Utility: KL divergence on trip distance / locations per user / aggregate time per location, home / work label error rate
Privacy: similarity between any synthetic trajectory and any real trajectory

Code:
https://github.com/aberke/lbs-data/tree/master/trajectory_synthesis

@misc{zhan2022privacyaware,
title={Privacy-Aware Human Mobility Prediction via Adversarial Networks},
author={Yuting Zhan and Alex Kyllo and Afra Mashhadi and Hamed Haddadi},
year={2022},
eprint={2201.07519},
archivePrefix={arXiv},
primaryClass={cs.LG}
}

name of the model: LSTM-PAE

DL components:
LSTM, adversarial learning

Evaluation metrics:
prediction accuracy (U), information loss in the reconstruction process (PI), user re-identification inaccuracy (PII)

• The BibTex of the paper:
  @Article{Arcolezi2022,
  doi = {10.1007/s00521-022-07393-0},
  url = {https://doi.org/10.1007/s00521-022-07393-0},
  year = {2022},
  month = jun,
  publisher = {Springer Science and Business Media {LLC}},
  author = {H{'{e}}ber Hwang Arcolezi and Jean-Fran{\c{c}}ois Couchot and Denis Renaud and Bechara Al Bouna and Xiaokui Xiao},
  title = {Differentially private multivariate time series forecasting of aggregated human mobility with deep learning: Input or gradient perturbation?},
  journal = {Neural Computing and Applications}
  }
• The name of the model proposed: No new model (evaluation of state-of-the-art shallow RNNs with differential privacy guarantees)
• The DL components adopted in the model: LSTM, GRU, BiLSTM, BiGRU
• The evaluation metrics adopted: RMSE and privacy-utility trade-off
• A link to an open datasets - https://github.com/hharcolezi/ldp-protocols-mobility-cdrs/blob/main/papers/%5B3%5D/ML_final_df_real.csv
• A link to a code repository - https://github.com/hharcolezi/ldp-protocols-mobility-cdrs/tree/main/papers/%5B3%5D

For a dataset

• The BibTex of the publication or the link related to the datasets: Same as before
• Dataset type: multivariate time series (6 coarse regions by 30-min time interval)
• The spatial and temporal coverage: Paris, France from Aug. 2020 to Nov. 2020
• The number of subjects covered: On average, 100k unique users per time interval
• The accessibility conditions: Fully open, with proper reference.
• A link to the dataset: https://github.com/hharcolezi/ldp-protocols-mobility-cdrs/blob/main/papers/%5B3%5D/ML_final_df_real.csv

In the tables with the papers, add a column "DOI" with the link to the paper.