Privacy-Preserving Logistic Regression with Distributed Data Sources via Homomorphic Encryption

Abstract

SUMMARYLogistic regression is a powerful machine learning tool to classify data. When dealing with sensitive or private data, cares are necessary. In this paper, we propose a secure system for privacy-protecting both the training and predicting data in logistic regression via homomorphic encryption. Perhaps surprisingly, despite the non-polynomial tasks of training and predicting in logistic regression, we show that only additively homomorphic encryption is needed to build our system. Indeed, we instantiate our system with Paillier, LWE-based, and ring-LWE-based encryption schemes, highlighting the merits and demerits of each instantiation. Besides examining the costs of computation and communication, we carefully test our system over real datasets to demonstrate its utility.

Characterization of differentially private logistic regression

“…In section 3.1, the supervised learning module and the feature extraction module of the proposed framework were evaluated using the subspace (ɑ3, ɑ4, ɑ11) of the NSL-KDD data set with binary classes (9,10) only. In a new experiment, the same subspace is again considered; however, the other classes (0,1), (0,5), (1,2), (1,9), (3,5), (3,9), and (6,8) are also studied. In addition, three other subspaces, (ɑ3, ɑ4, ɑ5), (ɑ3, ɑ4, ɑ7), and (ɑ4, ɑ7, ɑ10), are also included in the experiment to study the performance of the proposed analytical framework with DPLR.…”

“…The second column lists the maximum classification accuracies recorded for each classification task. For example, DPLR classifies the classes (1,9) with about 99% accuracy, whereas it classifies the classes (3,5) with about 53% accuracy. Since, DPLR's classification performance is poor for the binary class (3,5), classification accuracies are obtained for other subspaces, (ɑ3, ɑ4, ɑ5), (ɑ3, ɑ4, ɑ7), and (ɑ4, ɑ7, ɑ10) other subspaces as well, and listed in the third, fourth, and fifth columns of the table.…”

“…These characteristics of ϵ make the selection of a set of suitable values for ϵ much harder. A significant research has been performed to address this problem in various perspectives [1,2]; however, two main contributions that are closely related to the proposed study are selected and discussed in this section. For example, the authors of [7] studied the contributions of the privacy parameter and proposed a model that provides a balance between the objectives of a data owner and a data user, and studied its effectiveness by selecting several values of privacy parameter.…”

Input and Output Privacy-Preserving Linear Regression

“…Linear regression (whose outputs are continuous) is related but different from logistic regression (whose outputs are discrete). Various privacy-preserving systems for logistic regression have been built in [5]- [7], [19].…”

Privacy-Preserving SVM Computing by Using Random Unitary Transformation

“…For securing data, full encryption with provable security (like RSA, AES, etc) is the most secure option. However, many multimedia applications have been seeking a trade-off in security to enable other requirements, e.g., low processing demands, retaining bitstream compliance, and flexible processing in the encrypted domain, so that a lot of perceptual encryption schemes have been studied as one of the schemes for achieving a trade-off [4]- [13] In the recent years, considerable efforts have been made in the fields of fully homomorphic encryption and multi-party computation [14]- [17]. However, these schemes can not be applied yet to SVM algorithms, although it is possible to carry out some statistical analysis of categorical and ordinal data.…”