Syft 0.5: A Platform for Universally Deployable Structured Transparency

TL;DR

Syft 0.5 is a versatile framework integrating privacy technologies to enable structured transparency, demonstrated through a privacy-preserving neural network inference method that balances efficiency and model secrecy.

Contribution

Introduces a novel privacy-preserving inference flow using homomorphic encryption and model splitting, enhancing transparency and efficiency in neural network deployment.

Findings

Splitting the neural network reduces inference time.

Payload size of activation signals is decreased.

Trade-off between model secrecy and computational efficiency.

Abstract

We present Syft 0.5, a general-purpose framework that combines a core group of privacy-enhancing technologies that facilitate a universal set of structured transparency systems. This framework is demonstrated through the design and implementation of a novel privacy-preserving inference information flow where we pass homomorphically encrypted activation signals through a split neural network for inference. We show that splitting the model further up the computation chain significantly reduces the computation time of inference and the payload size of activation signals at the cost of model secrecy. We evaluate our proposed flow with respect to its provision of the core structural transparency principles.