Utility and Privacy Assessments of Synthetic Data for Regression Tasks

Abstract

With ever increasing capacity for collecting, storing, and processing of data, there is also a high demand for intelligent data analysis methods. While there have been impressive advances in machine learning and similar domains in recent years, this also gives rise to concerns regarding the protection of personal and otherwise sensitive data, especially if it is to be analysed by third parties. Besides anonymisation, which becomes challenging with high dimensional data, one approach for privacy-preserving data mining lies in the usage of synthetic data, which comes with the promise of protecting the users' data and producing analysis results close to those achieved by using real data. In this paper, we analyse a number of different approaches for creating synthetic data, and study the utility of the created datasets for regression tasks, i.e. the prediction of a numeric value. We further investigate the similarity of real and synthetic data samples. Finally, we contribute to privacy assessments and measurements of the risk of attribute disclosure on synthetic data by extending an approach developed for categorical data.

A Multi-Dimensional Evaluation of Synthetic Data Generators

“…Narrow or specific measures are widely used for assessing synthetic data [15], [19], [20], [27], [31], [32]. They are useful when the analysis to be performed on the synthetic data is known ahead of time.…”

“…Pairwise Correlations are sometimes measured using pairwise correlation plots such as heat maps [27], [32], but more often using statistical measures such as pairwise correlation difference (𝑃𝐶𝐷) [18]. We assess the correlations between attribute pairs using the latter.…”

Generating Synthetic Training Data for Supervised De-Identification of Electronic Health Records

“…A way to ensure this from a mathematical perspective is to train the generative models with a differential privacy (DP) objective. The premise of DP is that no output could be directly attributed to a single training instance [2,7,19,35]. In this study, we consciously chose not to include DP to maximize the utility of the synthetic corpora for the downstream task, but we recommend that future research uses DP in order to minimize privacy risks.…”

Optimasi Algoritma K-Means Clustering dengan Parallel Processing menggunakan Framework R

“…Dataset yang digunakan pada penelitian ini adalah data Boston Housing, yaitu data mengenai housing market di kota Boston, Amerika Serikat yang dikumpulkan oleh Statlib Library of Carnegie Mellon University [5]. Dataset ini sering dipakai pada penelitian mengenai data mining seperti pada penelitian prediksi housing prices [6] dan regression [7]. Algoritma K-Means diimplementasikan pada framework R menggunakan library.…”