Flash：一种面向深度 CNN、在 CPU 上兼具高精度与低延迟的混合隐私推理协议
Flash: A Hybrid Private Inference Protocol for Deep CNNs with High Accuracy and Low Latency on CPU
AI总结：
提出融合 HE 与 2PC 的 Flash 协议，通过快速卷积、多项式 $x^2+x$ 替换 ReLU 及免离线通信 2PC 求值，在 CPU 上显著降低深度 CNN 隐私推理延迟与通信成本。
摘要：
本文提出 Flash，这是一种同时利用同态加密（HE）和安全两方计算（2PC）的优化隐私推理（PI）混合协议，能够在 CPU 上将深度 CNN 模型的端到端 PI 延迟降至 1 分钟以内。为此，首先，Flash 提出一种基于快速槽旋转操作和新型数据编码方案的低延迟卷积算法，相比现有最优方法实现 4–94 倍的性能提升。其次，为最小化标准非线性激活函数 ReLU 引入的通信成本，Flash 使用多项式 $x^2+x$ 替换全部 ReLU，并采用新的训练策略训练深度 CNN 模型。与先前技术相比，训练后的模型在 CIFAR-10/100 和 TinyImageNet 上的推理准确率平均提升 16%（ResNet-32 最高提升 40%）。最后，Flash 提出一种高效的基于 2PC 的 $x^2+x$ 求值协议，该协议不需要任何离线通信，并将处理激活层的总通信成本相比现有最优方法降低 84–196 倍。因此，在 CPU 上实现的 Flash，其端到端 PI 延迟在 CIFAR-100 分类上为 0.02 分钟，在 TinyImageNet 分类上为 0.57 分钟，而总数据通信量在 CIFAR-100 上为 0.07GB，在 TinyImageNet 上为 0.22GB。Flash 将现有最优 PI 的延迟改进 16–45 倍，并将通信成本降低 84–196 倍。此外，即使对于 ImageNet，Flash 也能在 CPU 上实现低于 1 分钟的延迟，且总通信量低于 1GB。
英文摘要：
This paper presents Flash, an optimized private inference (PI) hybrid protocol utilizing both homomorphic encryption (HE) and secure two-party computation (2PC), which can reduce the end-to-end PI latency for deep CNN models less than 1 minute with CPU. To this end, first, Flash proposes a low-latency convolution algorithm built upon a fast slot rotation operation and a novel data encoding scheme, which results in 4-94x performance gain over the state-of-the-art. Second, to minimize the communication cost introduced by the standard nonlinear activation function ReLU, Flash replaces the entire ReLUs with the polynomial $x^2+x$ and trains deep CNN models with the new training strategy. The trained models improve the inference accuracy for CIFAR-10/100 and TinyImageNet by 16% on average (up to 40% for ResNet-32) compared to prior art. Last, Flash proposes an efficient 2PC-based $x^2+x$ evaluation protocol that does not require any offline communication and that reduces the total communication cost to process the activation layer by 84-196x over the state-of-the-art. As a result, the end-to-end PI latency of Flash implemented on CPU is 0.02 minute for CIFAR-100 and 0.57 minute for TinyImageNet classification, while the total data communication is 0.07GB for CIFAR-100 and 0.22GB for TinyImageNet. Flash improves the state-of-the-art PI by 16-45x in latency and 84-196x in communication cost. Moreover, even for ImageNet, Flash can deliver the latency less than 1 minute on CPU with the total communication less than 1GB.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
