Azure confidential computing instances offer the opportunity to quickly protect any application from insider threats, leveraging Intel® Software Guard Extensions (SGX) enabled CPUs and Anjuna Enterprise Enclaves software. With a single command, Anjuna automatically creates a secure enclave that isolates and encrypts all application resources in runtime, at rest, and on the network, achieving the strongest end-to-end data protection available; No changes to the application code or SDK required.
TensorFlow is a powerful end-to-end open-source platform for machine learning. It consists of an ecosystem of tools, libraries, and resources that lets subject matter experts build machine learning models and applications that can be deployed on cloud infrastructure.
These machine learning models are valuable and proprietary. Like all applications data, they are exposed in the cloud and vulnerable to theft. An insider can, for example, scan memory or storage using public scanning tools to gain easy access to the ML system, enabling exfiltration, unauthorized distribution, and unlicensed use. Protecting this IP is often critical to both businesses and researchers alike.
The Anjuna Enterprise Enclave for TensorFlow protects TensorFlow machine-learning models and intellectual property, including data and algorithms, from unlicensed and unauthorized use or theft. It leverages Azure’s confidential computing instances and Intel® Software Guard Extensions (SGX) to secure models placed in memory, storage, or sent over the network with hardware-grade encryption. Bad actors who gain access to Anjuna protected data, including these models, will not be able to see or make use of any application data--including the application itself.