Robust Privacy Preserving Linear Regression Using Beta-Divergence
-
- TAKESHITA Kotaro
- Tokyo Metropolitan College of Industrial Technology
-
- FUKUNAGA Shuichi
- Tokyo Metropolitan College of Industrial Technology
-
- TANAKA Satoru
- Tokyo Metropolitan College of Industrial Technology
-
- HUANG Xuping
- Advanced Institute of Industrial Technology
Bibliographic Information
- Other Title
-
- プライバシ保護機能をもつベータダイバージェンスを用いた ロバスト線形回帰
Abstract
This paper proposes a robust privacy preserving linear regression using beta-divergence. The proposed method encrypts sensitive data using an additive homomorphic encryption to achieve privacy preservation. As in the additive homomorphic encryption, the value of the data to be encrypted must be an integer. Although a weight function in a robust linear regression removes outliers, the output value of the weight function is a real number. The proposed method approximates the weight function as a polynomial, such that the output value of the approximated weight function is an integer. Numerical simulations demonstrate the effectiveness of the proposed method.
Journal
-
- 電子電子情報通信学会論文誌A 基礎・境界
-
電子電子情報通信学会論文誌A 基礎・境界 J105-A (6), 68-80, 2022-06-01
The Institute of Electronics, Information and Communication Engineers
- Tweet
Keywords
Details 詳細情報について
-
- CRID
- 1390292240175886848
-
- ISSN
- 18810195
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- KAKEN
-
- Abstract License Flag
- Disallowed