David Pointcheval

Théo Ryffel, Francis Bach, David Pointcheval

We analyse the privacy leakage of noisy stochastic gradient descent by modeling Rényi divergence dynamics with Langevin diffusions. Inspired by recent work on non-stochastic algorithms, we derive similar desirable properties in the stochastic setting. In particular, we prove that the privacy loss converges exponentially fast for smooth and strongly convex objectives under constant step size, which is a significant improvement over previous DP-SGD analyses. We also extend our analysis to arbitrary sequences of varying step sizes and derive new utility bounds. Last, we propose an implementation and our experiments show the practical utility of our approach compared to classical DP-SGD libraries.

Slim Bettaieb, Loic Bidoux, Olivier Blazy et al.

Decision forests are classical models to efficiently make decision on complex inputs with multiple features. While the global structure of the trees or forests is public, sensitive information have to be protected during the evaluation of some client inputs with respect to some server model. Indeed, the comparison thresholds on the server side may have economical value while the client inputs might be critical personal data. In addition, soundness is also important for the receiver. In our case, we will consider the server to be interested in the outcome of the model evaluation so that the client should not be able to bias it. In this paper, we propose a new offline/online protocol between a client and a server with a constant number of rounds in the online phase, with both privacy and soundness against malicious clients. CCS Concepts: $\bullet$ Security and Privacy $\rightarrow$ Cryptography.

Théo Ryffel, Pierre Tholoniat, David Pointcheval et al.

We propose AriaNN, a low-interaction privacy-preserving framework for private neural network training and inference on sensitive data. Our semi-honest 2-party computation protocol (with a trusted dealer) leverages function secret sharing, a recent lightweight cryptographic protocol that allows us to achieve an efficient online phase. We design optimized primitives for the building blocks of neural networks such as ReLU, MaxPool and BatchNorm. For instance, we perform private comparison for ReLU operations with a single message of the size of the input during the online phase, and with preprocessing keys close to 4X smaller than previous work. Last, we propose an extension to support n-party private federated learning. We implement our framework as an extensible system on top of PyTorch that leverages CPU and GPU hardware acceleration for cryptographic and machine learning operations. We evaluate our end-to-end system for private inference between distant servers on standard neural networks such as AlexNet, VGG16 or ResNet18, and for private training on smaller networks like LeNet. We show that computation rather than communication is the main bottleneck and that using GPUs together with reduced key size is a promising solution to overcome this barrier.

Theo Ryffel, Edouard Dufour-Sans, Romain Gay et al.

Machine learning on encrypted data has received a lot of attention thanks to recent breakthroughs in homomorphic encryption and secure multi-party computation. It allows outsourcing computation to untrusted servers without sacrificing privacy of sensitive data. We propose a practical framework to perform partially encrypted and privacy-preserving predictions which combines adversarial training and functional encryption. We first present a new functional encryption scheme to efficiently compute quadratic functions so that the data owner controls what can be computed but is not involved in the calculation: it provides a decryption key which allows one to learn a specific function evaluation of some encrypted data. We then show how to use it in machine learning to partially encrypt neural networks with quadratic activation functions at evaluation time, and we provide a thorough analysis of the information leaks based on indistinguishability of data items of the same label. Last, since most encryption schemes cannot deal with the last thresholding operation used for classification, we propose a training method to prevent selected sensitive features from leaking, which adversarially optimizes the network against an adversary trying to identify these features. This is interesting for several existing works using partially encrypted machine learning as it comes with little reduction on the model's accuracy and significantly improves data privacy.