| Project/Area Number |
17K00235
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Perceptual information processing
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2017-04-01 – 2021-03-31
|
| Project Status |
Completed (Fiscal Year 2020)
|
| Budget Amount *help |
¥4,680,000 (Direct Cost: ¥3,600,000、Indirect Cost: ¥1,080,000)
Fiscal Year 2020: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2019: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
Fiscal Year 2018: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2017: ¥780,000 (Direct Cost: ¥600,000、Indirect Cost: ¥180,000)
|
|---|
| Keywords | メディア認識 / パターン認識 / 情報セキュリティ / サーバ・クライアント / 認識器クローン / プライバシー保護 / Model Inversion Attack / クライアント・サーバ / 視覚メディア処理 |
|---|
| Outline of Final Research Achievements |
Client-server-based media recognition services, where client users send a media data to the recognition server while the server recognizes it and returns the result, have several risks for leaking sensitive information such as the recognition results, the server's recognition model, its training data, and so on. In this research project, we analyzed how much these risks are urgent and proposed some techniques to avoid or defend against them. The outcomes of this research project mainly include (i) media recognition framework where the recognition results are not disclosed to the server but correctly conveyed to the user, (ii) techniques to prevent and detect unauthorized clones of the server's recognition model, which we call "cloned recognizers", and (iii) techniques to estimate and regenerate a training data of a media recognition model only from the model itself.
|
| Academic Significance and Societal Importance of the Research Achievements |
AI技術の普及によりサーバ・クライアント型メディア認識サービスは既に現実のものとなりつつあり，今後の更なる発展が予想される中で，当該サービスを安心安全に運用・利用できないという事態になれば，大きな社会不安を引き起こす可能性が高い．本研究の成果は，そのリスクを低減するとともに，今後も継続して対処法の研究開発が求められることを示唆するものであり，安心安全なサービスの実現に大きく貢献し得る．また，学術的には，本研究の成果によりメディア認識分野・AI分野と情報セキュリティ分野を融合した新たな研究領域が創出される潜在性を持つ．他形態のメディア認識に対しても同様の研究を行う余地は大きく，極めて意義深い．
|
