Adaptive Laplace Mechanism: Differential Privacy Preservation in Deep Learning

Phan, NhatHai; Wu, Xintao; Han, Hu; Dou, Dejing

doi:10.1109/icdm.2017.48

Cited by 153 publications

(97 citation statements)

References 27 publications

(98 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are several common noise perturbation mechanisms for differential privacy that mask the original datasets or intermediate results during the training process of models: the Laplace mechanism, the exponential mechanism, and the Gaussian mechanism. Phan et al (2017) developed a novel mechanism that injects Laplace noise into the computation of Layer-Wise Relevance Propagation (LRP) to preserve differential privacy in deep learning. Chaudhuri et al (2011, 2013) adopted the exponential mechanism as a privacy-preserving tuning method by training classifiers with different parameters on disjoint subsets of the data and then randomizing the selection of which classifier to release.…”

Section: Differential Privacymentioning

confidence: 99%

A Privacy-Preserving Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition

Zhang

Xie

et al. 2020

Front. Neurorobot.

View full text Add to dashboard Cite

Recently, multi-task learning (MTL) has been extensively studied for various face processing tasks, including face detection, landmark localization, pose estimation, and gender recognition. This approach endeavors to train a better model by exploiting the synergy among the related tasks. However, the raw face dataset used for training often contains sensitive and private information, which can be maliciously recovered by carefully analyzing the model and outputs. To address this problem, we propose a novel privacy-preserving multi-task learning approach that utilizes the differential private stochastic gradient descent algorithm to optimize the end-to-end multi-task model and weighs the loss functions of multiple tasks to improve learning efficiency and prediction accuracy. Specifically, calibrated noise is added to the gradient of loss functions to preserve the privacy of the training data during model training. Furthermore, we exploit the homoscedastic uncertainty to balance different learning tasks. The experiments demonstrate that the proposed approach yields differential privacy guarantees without decreasing the accuracy of HyperFace under a desirable privacy budget.

show abstract

Section: Differential Privacymentioning

confidence: 99%

A Privacy-Preserving Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition

Zhang

Xie

et al. 2020

Front. Neurorobot.

View full text Add to dashboard Cite

show abstract

“…Their work was based on the prediction of human behavior in health social networks. [8] proposed an adaptive Laplace mechanism by combining differential privacy and layer-wise relevance propagation (LRP). The relevance obtained through the LRP process was altered with calculated noise to preserve privacy in the training model.…”

Section: Review Of Existing Privacy Preserving Modelsmentioning

confidence: 99%

“…The neuron contribution ratio is originally adapted from Layer-wise Relevance Propagation (LRP): an approach which has been proposed and adopted in literature [8,13,19,20]. The contribution ratio propagation involves a forward flow process that produces F θ at the output layer and a backward flow process that feeds back F θ as Crz into the network in a reversed manner, as presented in [20].…”

Section: Neuron Contribution Ratio Propagationmentioning

confidence: 99%

“…Recently, studies in [6][7][8][9][10] proposed some techniques for preserving privacy in deep neural networks. The techniques include fully homomorphic encryption (FHE) [8], (ϵ, δ)-differential privacy [11], and the adaptive Laplace mechanism [8] to mention just a few. Privacy preserving DNN models does exhibit lower accuracy as compared to non-privacy preserving counterparts due to enormous noise and lack of techniques to identify points to inject noise without denting accuracy and trading-off privacy.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Neuron Noise-Injection Technique for Privacy Preserving Deep Neural Networks

Adesuyi

Kim

2020

Open Computer Science

View full text Add to dashboard Cite

AbstractData is the key to information mining that unveils hidden knowledge. The ability to revealed knowledge relies on the extractable features of a dataset and likewise the depth of the mining model. Conversely, several of these datasets embed sensitive information that can engender privacy violation and are subsequently used to build deep neural network (DNN) models. Recent approaches to enact privacy and protect data sensitivity in DNN models does decline accuracy, thus, giving rise to significant accuracy disparity between a non-private DNN and a privacy preserving DNN model. This accuracy gap is due to the enormous uncalculated noise flooding and the inability to quantify the right level of noise required to perturb distinct neurons in the DNN model, hence, a dent in accuracy. Consequently, this has hindered the use of privacy protected DNN models in real life applications. In this paper, we present a neuron noise-injection technique based on layer-wise buffered contribution ratio forwarding and ϵ-differential privacy technique to preserve privacy in a DNN model. We adapt a layer-wise relevance propagation technique to compute contribution ratio for each neuron in our network at the pre-training phase. Based on the proportion of each neuron’s contribution ratio, we generate a noise-tuple via the Laplace mechanism, and this helps to eliminate unwanted noise flooding. The noise-tuple is subsequently injected into the training network through its neurons to preserve privacy of the training dataset in a differentially private manner. Hence, each neuron receives right proportion of noise as estimated via contribution ratio, and as a result, unquantifiable noise that drops accuracy of privacy preserving DNN models is avoided. Extensive experiments were conducted based on three real-world datasets and their results show that our approach was able to narrow down the existing accuracy gap to a close proximity, as well outperforms the state-of-the-art approaches in this context.

show abstract

“…Consider any belief distribution b. Let the posterior distributions b 0 [x i+S |y] and b i [x i+S |y] for some fixed i, S and y be defined in (9) and (15). From (19), ǫ-Bayesian differential privacy implies that for every z i+S ,…”

Section: Theorem 1 (Restated)mentioning

confidence: 99%

Relations among different privacy notions

Zhao¹

2017

2017 55th Annual Allerton Conference on Communication, Control, and Computing (Allerton)

View full text Add to dashboard Cite

We present a comprehensive view of the relations among several privacy notions: differential privacy (DP) [1], Bayesian differential privacy (BDP) [2], semantic privacy (SP) [3], and membership privacy (MP) [4]. The results are organized into two parts. In part one, we extend the notion of semantic privacy (SP) to Bayesian semantic privacy (BSP) and show its essential equivalence with Bayesian differential privacy (BDP) in the quantitative sense. We prove the relations between BDP, BSP, and SP as follows: ǫ-BDP ⇐= 1 2 − 1 e ǫ +1 -BSP, and ǫ-BDP =⇒ (e 2ǫ − 1)-BSP =⇒ (e 2ǫ − 1)-SP. In addition, we obtain a minor result ǫ-DP ⇐= 1 2 − 1 e ǫ +1 -SP, which improves the result of Kasiviswanathan and Smith [3] stating ǫ-DP ⇐= ǫ/6-SP for ǫ ≤ 1.35. In part two, we establish the relations between BDP and MP. First, ǫ-BDP =⇒ ǫ-MP. Second, for a family of distributions that are downward scalable in the sense of Li et al. [4], it is shown that ǫ-BDP ⇐= ǫ-MP.

show abstract

Adaptive Laplace Mechanism: Differential Privacy Preservation in Deep Learning

Cited by 153 publications

References 27 publications

A Privacy-Preserving Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition

A Privacy-Preserving Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition

A Neuron Noise-Injection Technique for Privacy Preserving Deep Neural Networks

Relations among different privacy notions

Contact Info

Product

Resources

About