| Presentation | 2021-07-16 A Study on Decentralized Machine Learning with Differential Privacy based on Input Perturbation Masakazu Okamoto, Koya Sato, Keiichi Iwamura, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Distributed machine learning eliminates the need for users to disclose their data to the out of the terminal since training can be done locally. However, machine learning has also been pointed out to have the potential to leak training data such as Model Inversion attack, which may lead to privacy violation. In this paper, we propose a method for satisfying differential privacy in distributed machine learning using input perturbations. Differential privacy is a definition for the privacy protection level, which can be satisfied by adding noise to the statistics. This allows us to analyze a large amount of data while ensuring privacy in distributed learning among users. Numerical simulations demonstrate the accuracy of the proposed and related methods. The results show that the proposed method can learn with higher accuracy than the output perturbation-based learning. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | distributed machine learning / differential privacy / input perturbation |
| Paper # | SR2021-34 |
| Date of Issue | 2021-07-07 (SR) |
| Conference Information | |
| Committee | RCS / SR / NS / SeMI / RCC |
|---|---|
| Conference Date | 2021/7/14(3days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Communication and Network Technology of the AI Age, M2M (Machine-to-Machine)，D2D (Device-to-Device)，IoT（Internet of Things）, etc |
| Chair | Eiji Okamoto(Nagoya Inst. of Tech.) / Suguru Kameda(Hiroshima Univ.) / Akihiro Nakao(Univ. of Tokyo) / Koji Yamamoto(Kyoto Univ.) / HUAN-BANG LI(NICT) |
| Vice Chair | Toshihiko Nishimura(Hokkaido Univ.) / Tomoya Tandai(Toshiba) / Fumihide Kojima(NICT) / Osamu Takyu(Shinshu Univ.) / Kentaro Ishidu(NICT) / Kazuto Yano(ATR) / Tetsuya Oishi(NTT) / Kazuya Monden(Hitachi) / Yasunori Owada(NICT) / Shunichi Azuma(Nagoya Univ.) / Koji Ishii(Kagawa Univ.) |
| Secretary | Toshihiko Nishimura(NEC) / Tomoya Tandai(Panasonic) / Fumihide Kojima(Mie Univ.) / Osamu Takyu(Tokai Univ.) / Kentaro Ishidu(NTT) / Kazuto Yano(NTT) / Tetsuya Oishi(Chuo Univ.) / Kazuya Monden(Cyber Univ.) / Yasunori Owada(Waseda Univ.) / Shunichi Azuma(Osaka Univ.) / Koji Ishii(CRIEPI) |
| Assistant | Koichi Adachi(Univ. of Electro-Comm.) / Osamu Nakamura(Sharp) / Manabu Sakai(Mitsubishi Electric) / Masashi Iwabuchi(NTT) / Tatsuki Okuyama(NTT DOCOMO) / Mai Ohta(Fukuoka Univ.) / Taichi Ohtsuji(NEC) / WANG Xiaoyan(Ibaraki Univ.) / Akemi Tanaka(MathWorks) / Kotaro Mihara(NTT) / Yuki Katsumata(NTT DOCOMO) / Akihito Taya(Aoyama Gakuin Univ.) / Yu Nakayama(Tokyo Univ. of Agri. and Tech.) / SHAN LIN(NICT) / Masaki Ogura(Osaka Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Radio Communication Systems / Technical Committee on Smart Radio / Technical Committee on Network Systems / Technical Committee on Sensor Network and Mobile Intelligence / Technical Committee on Reliable Communication and Control |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Study on Decentralized Machine Learning with Differential Privacy based on Input Perturbation |
| Sub Title (in English) | |
| Keyword(1) | distributed machine learning |
| Keyword(2) | differential privacy |
| Keyword(3) | input perturbation |
| 1st Author's Name | Masakazu Okamoto |
| 1st Author's Affiliation | Tokyo University of Science(Tokyo Univ. of Science) |
| 2nd Author's Name | Koya Sato |
| 2nd Author's Affiliation | Tokyo University of Science(Tokyo Univ. of Science) |
| 3rd Author's Name | Keiichi Iwamura |
| 3rd Author's Affiliation | Tokyo University of Science(Tokyo Univ. of Science) |
| Date | 2021-07-16 |
| Paper # | SR2021-34 |
| Volume (vol) | vol.121 |
| Number (no) | SR-104 |
| Page | pp.pp.67-72(SR), |
| #Pages | 6 |
| Date of Issue | 2021-07-07 (SR) |