Fast Homomorphic Evaluation of Deep Discretized Neural Networks

By Florian Bourse, Michele Minelli, Matthias Minihold, Pascal Paillier. 2018

358 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Efficiency & Compression, Privacy & Security

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Efficiency & Compression

In Privacy & Security

Citations in the corpus

Advances and Open Problems in Federated Learning (2019, methodology): “A common approach has been to adapt ML models and their training procedures to ensure that (over)underflows are controlled, by operating on normalized quantities and relying on careful quantization [172, 14, 182, 77].”
Secure Evaluation of Quantized Neural Networks (2019, methodology): “[8], where the authors use a quantization technique that is similar to the one described by Courbariaux and Bengio [13].”
QUOTIENT: Two-Party Secure Neural Network Training and Prediction (2019, methodology): “While there have been massive practical developments in both cryptography and ML (including the works above), most recent works for model training [22, 40, 41, 54] and prediction [7, 15, 50] on encrypted data are largely based on optimizations for either the ML model or the employed cryptographic te”
Security and Privacy Issues in Deep Learning (2018, methodology): “FHE-DiNN [19] utilized discretized neural networks with different weights and input dimensions to evaluate Chillotti et al.”
A Survey on Federated Learning Systems: Vision, Hype and Reality for Data Privacy and Protection (2019, methodology): “[179]. Here we introduce two major approaches that are adopted in current FLSs for data protection: cryptographic methods and differential privacy. Cryptographic methods such as homomorphic encryption[15, 77, 27, 33, 74, 145, 146, 204, 206, 112], and secure multi-party computation (SMC) [157, 36, 23”
Low Latency Privacy Preserving Inference (2018, methodology): “Bourse et al. (2017) used a different HE scheme that allows fast bootstrapping which results in only linear penalty for additional layers in the network.”
CHET: Compiler and Runtime for Homomorphic Evaluation of Tensor Programs (2018, methodology): “[9] use the TFHE library [17] for DNN inference.”
Secure Outsourced Matrix Computation and Application to Neural Networks (2018, neutral): “A number of subsequent works have improved it by normalizing weighted sums prior to applying the approximate activation function [10], or by employing a fully HE to apply an evaluation of an arbitrary deep neural networks [7].”
Faster CryptoNets: Leveraging Sparsity for Real-World Encrypted Inference (2018, background): “…[49] suggests that there are very signiﬁcant drawbacks to batching, including having to select more numerous and on accelerating neural networks neglect batching for similar reasons as ours ( [59] does not use batching, and [11]. uses batching to compress messages but not to improve throughput).”
TAPAS: Tricks to Accelerate (encrypted) Prediction As a Service (2018, methodology): “Most similar to our work is Bourse et al. (2017) who use neural networks with signed integer weights and binary activations to perform encrypted prediction.”