Falcon: Fast Spectral Inference on Encrypted Data

Authors: Qian Lou, Wen-jie Lu, Cheng Hong, Lei Jiang

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results show Falcon achieves the state-of-the-art inference accuracy and reduces the inference latency by 45.45% 85.34% over prior HENNs on MNIST and CIFAR-10.
Researcher Affiliation Collaboration Qian Lou Indiana University Bloomington louqian@iu.edu; Wen-jie Lu Alibaba Group juhou.lwj@alibaba-inc.com; Cheng Hong Alibaba Group vince.hc@alibaba-inc.com; Lei Jiang Indiana University Bloomington jiang60@iu.edu
Pseudocode Yes Algorithm 1: Homomorphic DFT (HDFT).; Algorithm 2: Homomorphic FC Layer (HFC).; Algorithm 3: Homomorphic Convolutional Layer.
Open Source Code No The paper does not include an unambiguous statement about releasing the source code for the methodology described, nor does it provide a direct link to a code repository.
Open Datasets Yes Our datasets include MNIST [18] and CIFAR-10 [19].; [18] Yann Le Cun, Corinna Cortes, and CJ Burges. MNIST Handwritten Digit Database. AT&T Labs [Online]. Available: http://yann.lecun.com/exdb/mnist, 2010.; [19] Alex Krizhevsky, Vinod Nair, and Geoffrey Hinton. The cifar-10 dataset, 2014. http://www.cs.toronto.edu/kriz/cifar.html.
Dataset Splits No The paper mentions using MNIST and CIFAR-10 datasets, but it does not explicitly provide specific details on training, validation, or test dataset splits (e.g., percentages, sample counts, or direct citations to predefined splits used for reproduction).
Hardware Specification Yes We ran all experiments on the same Azure standard B8ms virtual machine with 8 v CPUs and 32GB DRAM.
Software Dependencies Yes We use BFV scheme in SEAL [21] to implement Falcon. [21] Microsoft SEAL (release 3.4).
Experiment Setup Yes For MNIST and CIFAR-10, the plaintext modulus t = 2148728833 2148794369 2149810177, modulus degree N = 16384, coefficient modulus Q = 440 bits. More specific encryption parameters settings are shown in Table 7 and Table 8. The block size k of circulant matrix is set as 8 so that accuracy is not decreased. To keep original accuracy, the block size k = 16 of circulant matrix is used.