Differentially Private Individual Computation via Shuffling

Beyond Statistical Estimation: Differentially Private Individual Computation via Shuffling

This repository contains implementations of various privacy-preserving algorithms in the PIC model (tested with Python 3.10). The projects included are:

1. Spatial Crowdsourcing in the PIC Model, refer to `SpatialCrowdsourcing`.
2. Location-based Nearest Neighbors in the PIC Model, refer to `LocationBasedSocialSystem`.
3. Federated Learning with Incentives in the PIC Model, refer to `FederatedLearningIncentives`

## Setup

To get started with the projects, you need to go to sub-directory and install the application-specific dependencies. You can do this by running the following command:

```bash
pip install -r requirements.txt
```

Notice: This repository includes code and datasets from third parties for experimental performance comparisons. Please adhere to their respective licenses accordingly. The CC BY-NC-AD 4.0 International license applies to all other proprietary artifacts created by us.

Version 1.3: Provide an alternative and tighter way to privacy amplification via shuffling. More memory is required.
Version 1.2: Added correspondances between the parameter setting in READMEs and the experimentals results in the paper.
Version 1.1: Bug fix about declaring kNoisy in LocationBasedSocialSystem/nerestNeighbors.py.