Adaptive laplace mechanism: Differential privacy preservation in deep learning
Document Type
Conference Proceeding
Publication Date
12-15-2017
Abstract
In this paper, we focus on developing a novel mechanism to preserve differential privacy in deep neural networks, such that: (1) The privacy budget consumption is totally independent of the number of training steps; (2) It has the ability to adaptively inject noise into features based on the contribution of each to the output; and (3) It could be applied in a variety of different deep neural networks. To achieve this, we figure out a way to perturb affine transformations of neurons, and loss functions used in deep neural networks. In addition, our mechanism intentionally adds 'more noise' into features which are 'less relevant' to the model output, and vice-versa. Our theoretical analysis further derives the sensitivities and error bounds of our mechanism. Rigorous experiments conducted on MNIST and CIFAR-10 datasets show that our mechanism is highly effective and outperforms existing solutions.
Identifier
85043981824 (Scopus)
ISBN
[9781538638347]
Publication Title
Proceedings IEEE International Conference on Data Mining Icdm
External Full Text Location
https://doi.org/10.1109/ICDM.2017.48
ISSN
15504786
First Page
385
Last Page
394
Volume
2017-November
Grant
R01GM103309
Fund Ref
National Institutes of Health
Recommended Citation
Phan, Nhathai; Wu, Xintao; Hu, Han; and Dou, Dejing, "Adaptive laplace mechanism: Differential privacy preservation in deep learning" (2017). Faculty Publications. 9122.
https://digitalcommons.njit.edu/fac_pubs/9122
