Google Brain Adds Privacy to TensorFlow
- By John K. Waters
Google's deep learning and AI research team, Google Brain, has added yet another module to TensorFlow, the popular open source machine learning platform. Coming hot on the heels of the much anticipated TensorFlow 2.0 alpha release, TensorFlow Privacy is an open source library designed to make it easier for developers to train machine-learning models with privacy.
The new module addresses the growing challenges associated with training machine learning models on privacy-sensitive datasets -- things like personal photos and email. The module uses the theory of differential privacy, a statistical technique developed by cryptographers to maximize the accuracy of queries from a databases while minimizing the privacy impact on individuals whose information is in the database.
The Google Brain team is also billing the new module as an aid to researchers "by advancing the state of the art in machine learning with strong privacy guarantees." And it hopes the module will "develop into a hub of best-of-breed techniques for training machine-learning models with strong privacy guarantees," the group said in a blog post.
The Google Brain team has published an updated technical whitepaper ("A General Approach to Adding Differential Privacy to Iterative Training Procedures") describing its privacy mechanisms in detail. It also offered some examples of how this privacy mechanism might work, along with instructions for using it, in that blog post. The emphasis in this announcement is how easy it is to implement:
"To use TensorFlow Privacy, no expertise in privacy or its underlying mathematics should be required: those using standard TensorFlow mechanisms should not have to change their model architectures, training procedures, or processes. Instead, to train models that protect privacy for their training data, it is often sufficient for you to make some simple code changes and tune the hyperparameters relevant to privacy."
John has been covering the high-tech beat from Silicon Valley and the San Francisco Bay Area for nearly two decades. He serves as Editor-at-Large for Application Development Trends (www.ADTMag.com) and contributes regularly to Redmond Magazine, The Technology Horizons in Education Journal, and Campus Technology. He is the author of more than a dozen books, including The Everything Guide to Social Media; The Everything Computer Book; Blobitecture: Waveform Architecture and Digital Design; John Chambers and the Cisco Way; and Diablo: The Official Strategy Guide.