Encrypted Model

Encrypted models aim to perform machine learning computations on sensitive data without revealing the data itself, addressing crucial privacy concerns. Current research focuses on adapting various architectures, including transformers and wavelet neural networks, to work efficiently with homomorphic encryption and other cryptographic techniques like function secret sharing and federated learning with coded gradients. This field is significant because it enables the development of privacy-preserving machine learning applications across diverse sectors, ranging from personalized language models to secure collaborative training in sensitive domains like healthcare and finance.