| Item type |
Symposium(1) |
| 公開日 |
2022-10-17 |
| タイトル |
|
|
タイトル |
Risk Evaluation of LDP scheme LoPub against Variational Autoencoder |
| タイトル |
|
|
言語 |
en |
|
タイトル |
Risk Evaluation of LDP scheme LoPub against Variational Autoencoder |
| 言語 |
|
|
言語 |
eng |
| キーワード |
|
|
主題Scheme |
Other |
|
主題 |
Local Differential Privacy, Variational Auto-Encoder, LASSO regression |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_5794 |
|
資源タイプ |
conference paper |
| 著者所属 |
|
|
|
Meiji University |
| 著者所属 |
|
|
|
Meiji University |
| 著者所属(英) |
|
|
|
en |
|
|
Meiji University |
| 著者所属(英) |
|
|
|
en |
|
|
Meiji University |
| 著者名 |
Hernandez-Matamoros, Andres
Kikuchi, Hiroaki
|
| 著者名(英) |
Andres, Hernandez-Matamoros
Hiroaki, Kikuchi
|
| 論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
When we use a service which has an internet connection or not, the concern of how we can protect our data appears. Besides, the company that offers services wants to use our information to improve the service. Local Differential Privacy is applied to satisfy users' and companies' requirements about privacy concerns. This paper shows a risk Evaluation of LoPub, an LDP scheme. We show the results when the Central Server uses VAE instead uses the LoPub proposal. The results show that only a single VAE model performs well in a dataset with low cardinality on the attributes. Furthermore, it clears the way to continue researching VAE in LDP so that VAE can be applied to high cardinality datasets. |
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
When we use a service which has an internet connection or not, the concern of how we can protect our data appears. Besides, the company that offers services wants to use our information to improve the service. Local Differential Privacy is applied to satisfy users' and companies' requirements about privacy concerns. This paper shows a risk Evaluation of LoPub, an LDP scheme. We show the results when the Central Server uses VAE instead uses the LoPub proposal. The results show that only a single VAE model performs well in a dataset with low cardinality on the attributes. Furthermore, it clears the way to continue researching VAE in LDP so that VAE can be applied to high cardinality datasets. |
| 書誌情報 |
コンピュータセキュリティシンポジウム2022論文集
p. 289-296,
発行日 2022-10-17
|
| 出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |
IPSJ-CSS2022041.pdf (1.2 MB)
