Please visit [Google Scholar] for the latest publication list.
Privacy-preserving Machine Learning
In response to the growing demand for privacy and security, we delve into the realm of privacy-preserving machine learning. Our mission is to enhance the efficiency of deep learning models while striking a delicate balance between their accuracy and cryptographic overheads. Specifically, we strive to unlock the full potential of private inference, where inferences are conducted on encrypted data using advanced cryptographic techniques.
Private Computation Accelerators
We are developing cutting-edge hardware solutions for different privacy-preserving computation techniques, such as Homomorphic Encryption / Ring Processing, Garbled Circuits, and Secret Sharing. By harnessing the power of specialized accelerators, we aim to bring privacy-preserving technologies closer to real-world applications.
Zero-Knowledge Proof Accelerators
We are also pioneering accelerator designs for zero-knowledge proof (ZKP) systems, which are rapidly gaining traction in blockchain, privacy-preserving computation, and secure verification. Our work focuses on optimizing core proof primitives such as the SumCheck, Number Theoretic Transform (NTT), Merkle-tree and Multi-scalar Multiplication (MSM), enabling efficient zkSNARK constructions. By co-designing cryptographic algorithms and specialized hardware, we aim to significantly reduce proving costs and make ZKP-based applications more practical at scale.