Description
Striking a balance between security, privacy, and performance is a challenge in machine learning applications. In this talk we will present BlindAI, an open-source confidential computing solution that harnesses Intel SGX enclaves to enable secure remote ML inference. Our solution effectively safeguards the confidentiality of both the model and user data while also ensuring the predictions’ integrity.
We will discuss the motivation behind BlindAI, how we factored in the specificities and constraints of Intel SGX at the design stage, and share the outcome of an independent security audit of our solution.
Speakers
Corentin Lauverjat, Mithril Security
