LZLutan Zhao

Lutan Zhao

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Lutan Zhao has a total of 54 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • OmniZK: A Versatile Accelerator Architecture for Zero-Knowledge Proofs

    by Zhengbang Yang, Lutan Zhao, Han Liu, Kai Li, Rui Hou

  • Thunder: Efficient Multi-node FHE Acceleration Framework via In-Transit Computation

    by Kai Li, Lutan Zhao, Qingyun Niu, Yinhang Zheng, Zhengbang Yang, Boyan Zhao, Lu Gao, Rui Hou

  • Open Access

    Sub-Millisecond Gate Bootstrapping

    by Chunling Chen, Zhihao Li, Qingyun Niu, Xianhui Lu, Ruida Wang, Lutan Zhao, Rui Hou

  • Open Access

    SwiftFL: Enabling Speculative Training for On-Device Federated Deep Learning

    by Yuhui Zhang, Guang Yan, Xin Zhang, Zimu Guo, Lutan Zhao, Jiangfeng Cao, Dan Meng, Rui Hou

  • Open Access

    A survey of optimization techniques for bootstrapping algorithms in FHE

    by Tianyu Li, Zhiwei Wang, Lutan Zhao, Ruida Wang, Qingyun Niu, Xianhui Lu, Dan Meng, Rui Hou

    Fully Homomorphic Encryption (FHE) enables arbitrary computation on encrypted data without decryption, making it a cornerstone of privacy-preserving outsourcing, such as cloud computing. However, homomorphic operations cause ciphertext noise to grow until decryption fails. The efficient solution is bootstrapping, which refreshes the noise in FHE ciphertexts to sustain arbitrary deep homomorphic evaluation. But in practice, bootstrapping consumes over 50% of total execution time, posing a serious obstacle to FHE adoption. This paper presents a systematic survey of FHE bootstrapping algorithms and their optimizations. We organize existing works into three main paradigms: word-wise bootstrapping for BGV, BFV, and CKKS schemes; bit-wise bootstrapping for FHEW and TFHE schemes; and hybrid bootstrapping, which leverages both word-wise schemes and bit-wise schemes. We analyze the evolution of crucial techniques, highlight latest advances in reducing latency, enhancing parallelism, and controlling noise growth, and compare the advantages and limitations of different schemes. Finally, we discuss emerging research trends.


  • Open Access

    Maverick: Rethinking TFHE Bootstrapping on GPUs via Algorithm-Hardware Co-Design

    by Zhiwei Wang, Haoqi He, Lutan Zhao, Qingyun Niu, Dan Meng, Rui Hou

  • Open Access

    ShiftPIR: An Efficient PIR System with Gravity Shifting from Client to Server

    by Zihan Wang, Lutan Zhao, Ming Luo, Zhiwei Wang, Haoqi He, Wenzhe Lv, Xuan Ding, Dan Meng, Rui Hou


  • Open Access

    Chameleon: An Efficient FHE Scheme Switching Acceleration on GPUs

    by Zhiwei Wang, Haoqi He, Lutan Zhao, Peinan Li, Zhihao Li, Dan Meng, Rui Hou

    Fully homomorphic encryption (FHE) enables direct computation on encrypted data, making it a crucial technology for privacy protection. However, FHE suffers from significant performance bottlenecks. In this context, GPU acceleration offers a promising solution to bridge the performance gap. Existing efforts primarily focus on single-class FHE schemes, which fail to meet the diverse requirements of data types and functions, prompting the development of hybrid multi-class FHE schemes. However, studies have yet to thoroughly investigate specific GPU optimizations for hybrid FHE schemes. In this article, we present an efficient GPU-based FHE scheme switching acceleration named Chameleon. First, we propose a scalable NTT acceleration design that adapts to larger CKKS polynomials and smaller TFHE polynomials. Specifically, Chameleon tackles synchronization issues by fusing stages to reduce synchronization, employing polynomial coefficient shuffling to minimize synchronization scale, and utilizing an SM-aware combination strategy to identify the optimal switching point. Second, Chameleon is the first to comprehensively analyze and optimize critical switching operations. It introduces CMux-level parallelization to accelerate LUT evaluation and a homomorphic rotation-free matrix-vector multiplication to improve repacking efficiency. Finally, Chameleon outperforms the state-of-the-art GPU implementations by 1.23× in CKKS HMUL and 1.15× in bootstrapping. It also achieves up to 4.87× and 1.51× speedups for TFHE bootstrapping compared to CPU and GPU versions, respectively, and delivers a 67.3× average speedup for scheme switching over CPU-based implementation.


  • An Efficient Speculative Federated Tree Learning System With a Lightweight NN-Based Predictor

    by Yuhui Zhang, Hong Liao, Lutan Zhao, Yuncong Shao, Zhihong Tian, XiaoFeng Wang, Dan Meng, Rui Hou

    Federated tree-based models are popular in many real-world applications owing to their high accuracy and good interpretability. However, the classical synchronous method causes inefficient federated tree-based model training due to tree node dependencies. Inspired by speculative execution techniques in modern high-performance processors, this paper proposes FTSeir, a novel and efficient speculative federated learning system. Instead of simply waiting, FTSeir optimistically predicts the outcome of the prior tree node. By resolving tree node dependencies with a neural network-based split point predictor, the training tasks of child tree nodes can be executed speculatively in advance via separate threads. This speculation enables cross-layer concurrent training, thus significantly reducing the waiting time. Furthermore, we propose an eager verification mechanism to promptly identify mispredictions, thereby reducing wasted computing resources. On a misprediction, an incomplete rollback is triggered for quick recovery by reusing the output of the mis-speculative training, which reduces computational requirements. We implement FTSeir and evaluate its efficiency in a real-world federated learning setting with six public datasets. Evaluation results demonstrate that FTSeir achieves up to 3.45× and 3.60× speedup over the state-of-the-art gradient boosted decision trees and random forests implementations, respectively.


  • Open Access

    Elastic MSM: A Fast, Elastic and Modular Preprocessing Technique for Multi-Scalar Multiplication Algorithm on GPUs

    by Xudong Zhu, Haoqi He, Zhengbang Yang, Yi Deng, Lutan Zhao, Rui Hou

    Zero-knowledge proof (ZKP) is a cryptographic primitive that enables a prover to convince a verifier that a statement is true, without revealing any other information beyond the correctness of the statement itself. Due to its powerful capabilities, its most practical type, called zero-knowledge Succinct Non-interactive ARgument of Knowledge (zkSNARK), has been widely deployed in various privacypreserving applications such as cryptocurrencies and verifiable computation. Although state-of-the-art zkSNARKs are highly efficient for the verifier, the computational overhead for the prover is still orders of magnitude too high to warrant use in many applications. This overhead arises from several time-consuming operations, including large-scale matrix-vector multiplication (MUL), number-theoretic transform (NTT), and especially the multi-scalar multiplication (MSM) which constitutes the largest proportion. Therefore, further efficiency improvements are needed.In this paper, we focus on comprehensive optimization of running time and storage space required by the MSM algorithm on GPUs. Specifically, we propose a novel, modular and adaptive parameter configuration technique—elastic MSM to enable us to adjust the scale of MSM according to our own wishes by performing a corresponding amount of preprocessing. This technique enables us to fully unleash the potential of various efficient parallel MSM algorithms. We have implemented and tested elastic MSM over three prevailing parallel Pippenger algorithms on GPUs. Across various preprocessing space limitations (across various MSM scales), our constructions achieve up to about 1.90×, 1.08× and 1.36× (2.58×, 1.39× and 1.91×) speedup versus three state-of-the-art parallel Pippenger algorithms on GPUs, respectively.From another perspective, elastic MSM could also be regarded as a preprocessing technique over the well-known Pippenger algorithm, which is modular and could be used to accelerate almost all the most advanced parallel Pippenger algorithms on GPUs. Meanwhile, elastic MSM provides an adaptive trade-off between the running time and the extra storage space needed by parallel Pippenger algorithms on GPUs. This is the first preprocessing technique to retain the improved MSM computation brought by preprocessing under varying storage space limitations. Specifically, across various preprocessing space limitations (across various MSM scales), our constructions achieve up to about 192× and 223× (159× and 174×) speedup versus two state-ofthe- art preprocessing parallel Pippenger algorithms on GPUs, respectively.