Sarus Research

• Grislain et al. 2021 - DP-XGBoost: Private Machine Learning at Scale
  This paper describes the XGBoost fork we developed to turn any training with XGBoost into a privacy-preserving one.
  Released open-source on Github.
  

• Leduc et al. 2022 - Compiling Python Programs into Differentially Private Ones
  ‍This paper describes our method to compile a python program into a differentially private one.
  Peer-reviewed, accepted and presented at PEPR 2022.

• Leduc et al. 2021 - Composable Generative Models
  ‍This paper shows how heterogeneous generative models can be combined into one.
  
• Canale et al. 2022 - Generative Modeling of Complex Data
  ‍This paper shows how complex hierarchical data-types composed of numbers, categories, text, pictures, series can be mapped to generative models of these data.
  Peer-reviewed, accepted and presented at PSD 2022.
  

• Lalanne et al. 2023 - Private Quantiles Estimation in the Presence of Atoms
  This paper describes a method to estimate quantiles with privacy guarantees.
  Peer-reviewed, accepted and published in Information and Inference: A Journal of the IMA.
  

• Grislain et al. 2023 - Qrlew: Rewriting SQL into Differentially Private SQL
  ‍This paper describes a method to turn a SQL query into a SQL query with Differential Privacy.
  Peer-reviewed, accepted and presented at AAAI - PPAI 2024.
  Released open source on Github.
  
• Grislain et al. 2024 - Qrlew : Transformer du SQL en SQL avec confidentialité différentielle
  ‍This paper describes a method to turn a SQL query into a SQL query with Differential Privacy.
  Peer-reviewed, accepted and presented at APVP 2024