Secure multi-party AI training with Confidential Computing

Multi-party learning is the key to accessing more data and developing more efficient AI. For instance, data is often scarce in healthcare and siloed in small datasets. But a high level of data protection is crucial to convince more organizations to collaborate. Several technologies, such as Federated Learning, have emerged. However, implementing them at scale and securely is still a real challenge. 

That is why Mithril Security is building BastionAI, an open-source privacy-friendly deep learning framework to train AI models at scale with secure enclaves. First benchmark shows that training an EfficientNet with BastionAI is 2 times to 20 times faster compared to training it with Pytorch Federated learning with Differential Privacy. Let’s meet on October 17th at the SCAI. We will discuss secure multi-party training, Differential Privacy, Federated Learning, Confidential Computing, and our latest product, BastionAI.

To register : https://www.eventbrite.fr/e/bastionai-secure-multi-party-ai-training-with-confidential-computing-tickets-430788539117?aff=ebdsoporgprofile