A Differential Privacy Support Vector Machine Classifier Based on Dual Variable Perturbation

Abstract

Data mining technology can be used to dig out potential and valuable information from massive data, and support vector machine (SVM) is one of the most widely used and most efficient methods in the field of data mining classification. However, the training set data often contains sensitive attributes, and the traditional training method of SVM reveals the individual privacy information. In view of the low prediction accuracy and poor versatility of the existing SVM classifiers with privacy protection, this paper proposed a new SVM training method for differential privacy protection. The algorithm first solved the dual problem of SVM by using SMO method and the difference E i between the estimated value and the real value for each support vector was recorded. Then the ratio of the E i of each support vector to the sum of the E i of all the support vectors was calculated. Next, different levels of Laplace random noise were added to the corresponding dual variables α i of each support vector to be released, according to the ratio of each support vector. According to the principle of differential privacy protection, the algorithm meets-differential privacy which can be used to effectively protect individual privacy. Experimental results on real datasets showed that the algorithm proposed in this paper could be used for classification prediction under a reasonable privacy budget.

DPWSS: differentially private working set selection for training support vector machines

“…The datasets are partly selected for the experiments as Zhang, Hao & Wang (2019) , Fan, Chen & Lin (2005) and Zhao et al (2007) . All datasets are for binary classification, and available at .…”

“…However, public data is hard to obtain in the modern private world. Zhang, Hao & Wang (2019) constructed a novel private SVM classifier by dual variable perturbation, which adds Laplace noise to the corresponding dual variables according to the ratio of errors.…”

Privacy preserving classification over differentially private data

“…The early version of differentially private SVM has been proposed by Rubinstein et al (2009). After a long time, Zhang et al (2019) have proposed a differentially private SVM based‐on dual variable perturbation. Recently, Senekane (2019) has performed differentially private image classification using SVM.…”

“…Recently, Senekane (2019) has performed differentially private image classification using SVM. For these differentially private SVM classification algorithms, Rubinstein et al (2009) and Zhang et al (2019) utilize output perturbation technique from differential privacy, while Senekane (2019) employs input perturbation technique.…”

“…In the literature, differentially private classification algorithms based on k-NN, ID3, random decision trees, and forests, Naïve Bayes, SVM, and Holte's One Rule have been proposed (Bojarski et al, 2014;Fletcher & Islam, 2015, 2019Friedman & Schuster, 2010;Gursoy et al, 2017;Rubinstein et al, 2009;Senekane, 2019;Vaidya, Shafiq, Basu, & Hong, 2013;Zhang, Hao, & Wang, 2019;Zorarpacı & Özel, 2020). Briefly overviewing the literature concerning classification with differential privacy, it can be seen that, the majority of the existing methods employ output perturbation technique from differential privacy.…”