hharcolezi / risks-ldp

Repository for the VLDB'23 paper "On the Risks of Collecting Multidimensional Data Under Local Differential Privacy"

On the Risks of Collecting Multidimensional Data Under LDP

Héber H. Arcolezi, Sébastien Gambs, Jean-François Couchot, Catuscia Palamidessi. "On the Risks of Collecting Multidimensional Data Under Local Differential Privacy". PVLDB, 16(5): 1126 - 1139, 2023. doi: 10.14778/3579075.3579086.

If our codes and work are useful to you, we would appreciate a reference to:

@article{Arcolezi2023,
  doi = {10.14778/3579075.3579086},
  url = {https://doi.org/10.14778/3579075.3579086},
  year = {2023},
  month = jan,
  publisher = {Association for Computing Machinery ({ACM})},
  volume = {16},
  number = {5},
  pages = {1126--1139},
  author = {H{\'{e}}ber H. Arcolezi and S{\'{e}}bastien Gambs and Jean-Fran{\c{c}}ois Couchot and Catuscia Palamidessi},
  title = {On the Risks of Collecting Multidimensional Data Under Local Differential Privacy},
  journal = {Proceedings of the {VLDB} Endowment}
}

Codes

• The attack_SMP folder has the codes for reproducing the attacks to the SMP solution.
• The attack_RSpFD folder has the codes for reproducing the attacks to the RS+FD solution.
• The countermeasure_RSpRFD folder has the codes for reproducing the experiments/attacks of our countermeasure RS+RFD solution.

Datasets

The datasets folder has the following (pre-processed) datasets.

• ACSEmployement
• Adult

To Do

• I am slowly cleaning/generalizing the codes + documentation.
• Implement RS+RFD in the multi-freq-ldpy package.

Environment

I mainly used Python 3 with numpy, pandas, numba, multi-freq-ldpy, and ray libaries. The versions I use are listed below:

• Python 3.8.8
• Numpy 1.23.1
• Pandas 1.2.4
• Numba 0.53.1
• Multi-freq-ldpy 0.2.4
• Ray 1.11.0

Contact

For any question, please contact Héber H. Arcolezi: heber.hwang-arcolezi [at] inria.fr

License

MIT