Laplace Input and Output Perturbation for Differentially Private Principal Components Analysis

Abstract

With the widespread application of big data, privacy-preserving data analysis has become a topic of increasing significance. The current research studies mainly focus on privacy-preserving classification and regression. However, principal component analysis (PCA) is also an effective data analysis method which can be used to reduce the data dimensionality, commonly used in data processing, machine learning, and data mining. In order to implement approximate PCA while preserving data privacy, we apply the Laplace mechanism to propose two differential privacy principal component analysis algorithms: Laplace input perturbation (LIP) and Laplace output perturbation (LOP). We evaluate the performance of LIP and LOP in terms of noise magnitude and approximation error theoretically and experimentally. In addition, we explore the variation of performance of the two algorithms with different parameters such as number of samples, target dimension, and privacy parameter. Theoretical and experimental results show that algorithm LIP adds less noise and has lower approximation error than LOP. To verify the effectiveness of algorithm LIP, we compare our LIP with other algorithms. The experimental results show that algorithm LIP can provide strong privacy guarantee and good data utility.

Beneficial influence of Hf and Zr additions to Nb4at%Ta on the vortex pinning of Nb3Sn with and without an O source

“…Nb4Ta in a PIT wire have also been recently reported by Xu et al [32]. However, in this comparison of properties with and without SnO2, we note that we do still see a small HIrr degradation in the presence of O, the effect being significantly worse in the Zr alloy.…”

Differential Privacy Principal Component Analysis for Support Vector Machines

“…e Laplace mechanism adds independent noise to data; we use lap(b) to represent the noise sampled from Laplace distribution with a scaling of b. Definition 3 (Laplace mechanism) (see [17]). Given a data set D, for a function f: D ⟶ R d , with sensitivity Δf, the mechanism M provides ε− differential privacy satisfying…”

The legacy of the experimental hadron physics programme at COSY

“…27, where they are compared with the results from ANL at 2.2 GeV [79]. The overall normalizations of the KOALA data have not yet been determined because the completely pure Coulomb region was not fully accessed and the analysis is still proceeding [83].…”

“…Elastic proton-proton differential cross sections at small momentum transfers t. The preliminary values measured by the PANDA collaboration with the KOALA detector at 2.5, 2.8, and 3.2 GeV/c [82] are compared to the angular dependence measured at ANL at 3.0 GeV/c [79]. The normalizations of the KOALA data at the three momenta are still under study [83] and all data are given in arbitrary units. 85].…”