CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy

By Nathan Dowlin, Ran Gilad-Bachrach, Kim Laine, Kristin E. Lauter, Michael Naehrig, John Robert Wernsing. 2016

1,950 citations recorded by the source corpus.

Subfields: Safety, Robustness & Interpretability

How researchers see this paper

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

In Safety, Robustness & Interpretability

Citations in the corpus

Towards the Science of Security and Privacy in Machine Learning (2016, background): “For instance, Dowlin et al. use homomorphic encryption [104] to encrypt the data in a form that allows a neural network to process it without decrypting it [105].”
Advances and Open Problems in Federated Learning (2019, background): “s. 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]. It has been known for several decades that any function can be securely computed”
Deep Models Under the GAN: Information Leakage from Collaborative Deep Learning (2017, methodology): “Microsoft developed CryptoNets [20] to perform deep learning on encrypted data and provide encrypted outputs to the users [86].”
Secure Evaluation of Quantized Neural Networks (2019, methodology): “One can map a set of ﬂoating-point numbers with ﬁnite precision to integers and use a modulus that is large enough so that overﬂows do not occur (effectively emulating integer arithmetic), as done by Gilad-Bachrach et al. [24].”
HEAX: An Architecture for Computing on Encrypted Data (2019, methodology): “For instance, in the context of Machine Learning as a Service (MLaaS), FHE can be used to perform oblivious neural network inference [25, 35]: clients send the encrypted version of their data, the cloud server runs ML models on the encrypted queries, and returns the result to the clients.”
Learning Privately over Distributed Features: An ADMM Sharing Approach (2019, background): “Another approach to privacy-preserving machine learning is through encryption Gilad-Bachrach et al. (2016); Takabi et al. (2016); Kikuchi et al. (2018) or secret sharing Mohassel and Zhang (2017); Wan et al. (2007); Bonte and Vercauteren (2018), so that models are trained on encrypted data.”
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”
Federated Machine Learning (2019, background): “e the privacy of gradients and enhance the security of the system. With the recent advances in deep learning, privacy-preserving neural networks inference is also receiving a lot of research interests[10,11,14,28,40,52,54]. ACM Trans. Intell. Syst. Technol., Vol. 10, No. 2, Article 12. Publication d”
Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware (2018, methodology): “, in the latter two where the computation is performed locally by the data owner), the need for integrity is a common denominator, and this is the property we primarily focus on in this paper (and for which we obtain the best results!) Cryptography can address some of these problems [17, 32, 3, 38, ”
Contamination Attacks and Mitigation in Multi-Party Machine Learning (2018, neutral): “Secure MPC can be enabled with cryptographic techniques [6, 11, 18, 29, 32], and systems based on trusted processors such as Intel SGX [3, 5, 33].”
The Value of Collaboration in Convex Machine Learning with Differential Privacy (2019, methodology): “In [15], efficient conversion of models for use of encrypted input data was discussed.”
Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims (2020, methodology): “The most well known methods for encrypted computation are homomorphic encryption, secure multi-party computation, and functional encryption [102].”
Learning with Selective Forgetting (2021, methodology): “This has brought new challenges to the ﬁeld, covering learning from encrypted data [Gilad-Bachrach et al. , 2016], preventing learning of unintended information [Wang et al. , 2019], privacy preserving localization [ Speciale et al. , 2019a; Speciale et al. , 2019b ] , just to name a few.”
DeepSecure: Scalable Provably-Secure Deep Learning (2017, methodology): “a, while ensuring the confidentiality of both data and DL parameters. To address DL execution in sensitive applications, Microsoft has recently proposed a cryptographic DL framework (called CryptoNet [8]) based on Homomorphic Encryption (HE). Although this approach enables DL execution in a privacy-”