Private Machine Learning in TensorFlow using Secure Computation
This work addresses the problem of implementing private machine learning for researchers and practitioners by integrating secure computation into a familiar framework, though it is incremental as it builds on existing protocols and tools.
The authors developed a framework for secure multi-party computation within TensorFlow, enabling private machine learning experiments with tight integration into existing workflows and leveraging TensorFlow's distributed computation optimizations. They provided an open-source implementation of a state-of-the-art protocol and reported concrete benchmarks using typical models.
We present a framework for experimenting with secure multi-party computation directly in TensorFlow. By doing so we benefit from several properties valuable to both researchers and practitioners, including tight integration with ordinary machine learning processes, existing optimizations for distributed computation in TensorFlow, high-level abstractions for expressing complex algorithms and protocols, and an expanded set of familiar tooling. We give an open source implementation of a state-of-the-art protocol and report on concrete benchmarks using typical models from private machine learning.