标题： 隐私保护的量化BERT模型推理 显示英文标题

标题： Privacy-Preserving Inference for Quantized BERT Models

摘要： 随着生成式机器学习模型在医疗保健和个人化服务等隐私敏感领域中的日益部署，确保安全推理已成为一个关键挑战。安全多方计算（MPC）能够实现隐私保护的模型推理，但其存在较高的通信和计算开销。主要瓶颈在于浮点运算的昂贵安全评估。量化通过将浮点运算转换为低精度整数计算，提供了一个有前景的解决方案，显著减少了开销。然而，现有的基于MPC的量化推理方法要么依赖于公开的量化参数——带来隐私风险，要么在处理非线性函数如激活函数和softmax时效率低下。在本工作中，我们提出了一种细粒度、逐层的量化方案，并在安全环境中支持1比特权重的全连接层。我们设计了一个多输入查找表协议，以高效且安全地评估softmax。此外，我们使用双秘密共享方案，并通过查找表进行精度转换，完全消除了截断开销。在BERT-base模型上的实验评估表明，我们的方法相比Lu\emph{等}的速度提升了$8\times$，相比Gupta\emph{等}的速度提升了$9\times$，相比Knott\emph{等}的速度提升了$22 \times$。 显示英文摘要

摘要： With the increasing deployment of generative machine learning models in privacy-sensitive domains such as healthcare and personalized services, ensuring secure inference has become a critical challenge. Secure multi-party computation (MPC) enables privacy-preserving model inference but suffers from high communication and computation overhead. The main bottleneck lies in the expensive secure evaluation of floating-point operations. Quantization offers a promising solution by converting floating-point operations into lower-precision integer computations, significantly reducing overhead. However, existing MPC-based quantized inference methods either rely on public quantization parameters-posing privacy risks-or suffer from inefficiencies, particularly in handling nonlinear functions such as activations and softmax. In this work, we propose a fine-grained, layer-wise quantization scheme and support 1-bit weight fully connected layers in a secure setting. We design a multi-input lookup table protocol to evaluate softmax efficiently and securely. Furthermore, we use dual secret sharing schemes and perform precision conversions via lookup tables, eliminating truncation overhead entirely. Experimental evaluation on BERT-base models demonstrates that our approach achieves up to $8\times$ speedup compared to Lu \emph{et al}. (NDSS 25), $9\times$ speedup compared to Gupta \emph{et al}. (PETS 24) and $22 \times$ speedup compared to Knott \emph{et al}. (NeurIPS 21).