https://github.com/ChrisWaites/pyvacy

Differentially Private Optimization for PyTorch 👁🙅‍♀️
https://github.com/ChrisWaites/pyvacy

differential-privacy pytorch

Last synced: 3 months ago
JSON representation

Differentially Private Optimization for PyTorch 👁🙅‍♀️

• Host: GitHub
• URL: https://github.com/ChrisWaites/pyvacy
• Owner: ChrisWaites
• License: apache-2.0
• Created: 2019-03-22T22:05:42.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2020-05-26T15:29:38.000Z (over 4 years ago)
• Last Synced: 2024-03-15T04:11:49.611Z (11 months ago)
• Topics: differential-privacy, pytorch
• Language: Python
• Homepage:
• Size: 1.38 MB
• Stars: 179
• Watchers: 6
• Forks: 29
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• AwesomeResponsibleAI - PyVacy: Privacy Algorithms for PyTorch

README

        


    


    
PyVacy: Privacy Algorithms for PyTorch



Basically [TensorFlow Privacy](https://github.com/tensorflow/privacy), but for PyTorch.

DP-SGD implementation modeled after techniques presented within [Deep Learning with Differential Privacy](https://arxiv.org/abs/1607.00133) and [A General Approach to Adding Differential Privacy to Iterative Training Procedures](https://arxiv.org/abs/1812.06210).

## Example Usage

```python

from pyvacy import optim, analysis, sampling

training_parameters = {

    'N': len(train_dataset),

    # An upper bound on the L2 norm of each gradient update.

    # A good rule of thumb is to use the median of the L2 norms observed

    # throughout a non-private training loop.

    'l2_norm_clip': 1.0,

    # A coefficient used to scale the standard deviation of the noise applied to gradients.

    'noise_multiplier': 1.1,

    # Each example is given probability of being selected with minibatch_size / N.

    # Hence this value is only the expected size of each minibatch, not the actual. 

    'minibatch_size': 128,

    # Each minibatch is partitioned into distinct groups of this size.

    # The smaller this value, the less noise that needs to be applied to achieve

    # the same privacy, and likely faster convergence. Although this will increase the runtime.

    'microbatch_size': 1,

    # The usual privacy parameter for (ε,δ)-Differential Privacy.

    # A generic selection for this value is 1/(N^1.1), but it's very application dependent.

    'delta': 1e-5,

    # The number of minibatches to process in the training loop.

    'iterations': 15000,

}

model = nn.Sequential(...)

optimizer = optim.DPSGD(params=model.parameters(), **training_parameters) 

epsilon = analysis.epsilon(**training_parameters)

loss_function = ...

minibatch_loader, microbatch_loader = sampling.get_data_loaders(**training_parameters)

for X_minibatch, y_minibatch in minibatch_loader(train_dataset):

    optimizer.zero_grad()

    for X_microbatch, y_microbatch in microbatch_loader(TensorDataset(X_minibatch, y_minibatch)):

        optimizer.zero_microbatch_grad()

        loss = loss_function(model(X_microbatch), y_microbatch)

        loss.backward()

        optimizer.microbatch_step()

    optimizer.step()

```

## Tutorials

`mnist.py`

Implements a basic classifier for identifying which digit a given MNIST image corresponds to. The model achieves a test set classification accuracy of 96.7%. The architecture and results achieved are inspired by the corresponding tutorial within [TensorFlow privacy](https://github.com/tensorflow/privacy/tree/master/tutorials).

## Disclaimer

Do NOT use the contents of this repository in applications which handle sensitive data. The author accepts no liability for privacy infringements - use the contents of this repository solely at your own discretion.