# tensorflow/privacy

Library for training machine learning models with privacy for training data

Repository: https://github.com/tensorflow/privacy
Canonical: https://ross.abutalabs.com/products/privacy
Language: Python
License: Apache-2.0
License Family: permissive
Topics: machine-learning, privacy
Last push: 2026-08-26T19:07:16+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 8, longevity 100
- inputs: {"age_days": 2812, "days_push": 7, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2026, forks 475 (observed 2026-08-28T04:06:06.872675+00:00)

## What it is
TensorFlow Privacy is a Python library providing TensorFlow optimizers for training machine learning models with differential privacy. It includes tutorials and analysis tools for computing the privacy guarantees of trained models.

## Use cases
- train a neural network with differential privacy guarantees
- add DP-SGD gradient clipping to a Keras model
- compute the epsilon privacy budget of a training run
- test models for empirical privacy leakage
- train models on sensitive data like medical records

## When to choose
- you train TensorFlow models on sensitive or personal data and need formal privacy guarantees
- you need membership inference analysis or privacy accounting tools
- you want DP optimizers that integrate with existing Keras workflows

## When to avoid
- you use PyTorch or JAX rather than TensorFlow (consider Opacus instead)
- you need privacy guarantees for inference or data release rather than model training
- you need production-grade DP training with minimal overhead on complex architectures

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, privacy, security, llm-training
- domain: machine-learning, privacy, deep-learning
- platform: python
- tags: differential-privacy, tensorflow, dp-sgd, gradient-clipping, privacy-accounting

## Member repositories
- tensorflow/privacy (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:06.872675+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:59:45.071128+00:00, confidence not recorded.
  - readme: https://github.com/tensorflow/privacy (fetched 2026-08-28T04:06:06.872675+00:00, sha 010d0782ead5)
- Data as of 2026-08-30T08:39:29.467469+00:00.