Latest Blog Posts
Latest News & Blog Posts
The Role of Red Teaming in PETs: In February, our team won the Red Teaming category of the U.S. PETs Prize Challenge, securing a prize of 60,000 USD. In this blog post, we will provide a brief overview of the significance of Red Teaming in the field of Privacy Enhancing Technologies (PETs) research in the context of the competition.
LAMP: Extracting text from gradients with language model priors: In this work we present an attack on federated learning's privacy specific to the text domain. We show that federated learning in the text domain can expose a lot of user data.
Reliability guarantees on private data: We present Phoenix (CCS '22), the first system for privacy-preserving neural network inference with robustness and fairness guarantees.
Why tighter convex relaxations harm certified training: We investigate a long-standing paradox in the field of certified training, identifying previously overlooked properties of convex relaxations which affect training success.