Private AI

Cachemir: Fully Homomorphic Encrypted Inference of Generative Large Language Model with KV Cache
A fully homomorphic encrypted LLM inference framework that efficiently integrates the KV cache for autoregressive decoding.
Ye Yu, Yifan Zhou, Yi Chen, Pedro Soto, Wenjie Xiong, Meng Li
Paper
ROSETTA: Efficient and Accurate Privacy-Preserving LLM Decoding via Hybrid CKKS/TFHE Evaluation
A hybrid CKKS/TFHE framework for efficient and accurate privacy-preserving autoregressive LLM decoding.
Jiangrui Yu, Baosheng Zhang, Liang Kong, Lin Ding, Yi Chen, Ye Yu, Mingzhe Zhang, Meng Li
Paper
Helios: Melting Kernel Boundaries for GPU-Accelerated HE via Graph Rewriting and Microarchitecture-Aware Mapping
Yi Chen, Ziyu Tang, Chao Yang, Guang Fan, Mingzhe Zhang, Meng Li
OptiPrime: Optimizing Private Inference through protocol-hardware codesign
Jiangrui Yu, Ye Yu, Si Chen, Chenqi Lin, Wenxuan Zeng, Junfeng Fan, Mingyu Gao, Meng Li
PEFT: A Near-Memory Processing-Enabled Heterogeneous Accelerator for BatchPBS TFHE
A near-memory processing-enabled heterogeneous accelerator that reduces data movement for BatchPBS TFHE workloads.
Paper
CryptoMoE: Privacy-Preserving and Scalable Mixture of Experts Inference via Balanced Expert Routing
Yifan Zhou, Tianshi Xu, Jue Hong, Ye Wu, Meng Li
Paper
MPCache: MPC-Friendly KV Cache Eviction for Efficient Private LLM Inference
Wenxuan Zeng, Ye Dong, Jinjin Zhou, Jin Tan, Lei Wang, Tao Wei, Runsheng Wang, Meng Li
Paper
FENIX: Flexible and Efficient Hybrid HE/MPC Acceleration with Near-Memory Processing
Tengyu Zhang, Chenqi Lin, Jiangrui Yu, Shuwen Deng, Meng Li
Paper
Ironman: Accelerating Oblivious Transfer Extension for Privacy-Preserving AI with Near-Memory Processing
A near-memory architecture that accelerates oblivious-transfer extension, a key bottleneck in privacy-preserving AI.
Chenqi Lin, Kang Yang, Tianshi Xu, Ling Liang, Yufei Wang, Zhaohui Chen, Runsheng Wang, Mingyu Gao, Meng Li
Paper
Breaking the Layer Barrier: Remodeling Private Transformer Inference with Hybrid CKKS and MPC
A hybrid CKKS/MPC framework that remodels private Transformer inference through fine-grained operator fusion.
Tianshi Xu, Wenjie Lu, Jiangrui Yu, Chenqi Lin, Yi Chen, Runsheng Wang, Meng Li
Paper