Intellectual Property Protection Oriented AI with Secure Authentication and Its Evaluation
-
- IKEZAKI Yoshiya
- Meijo University
-
- TAKEMOTO Shu
- Meijo University
-
- NOZAKI Yusuke
- Meijo University
-
- YOSHIKAWA Masaya
- Meijo University
Bibliographic Information
- Other Title
-
- 知財保護を指向したAIエッジ・認証併用システムとその評価
Abstract
It is necessary not only to authenticate devices but also to protect AI model information used in edge devices from the viewpoint of corporate IP protection. In this paper, we propose a new combined AI edge and authentication system oriented toward IP protection. The proposed method utilizes Physically Unclonable Function (PUF) to prevent from using unauthorized AI models. Experiments show that the proposed method successfully reduced the inference accuracy by more than 10% for an illegally duplicated AI device, even though the circuit configuration is identical to the original device.
Journal
-
- 電子情報通信学会論文誌D 情報・システム
-
電子情報通信学会論文誌D 情報・システム J106-D (4), 209-217, 2023-04-01
The Institute of Electronics, Information and Communication Engineers
- Tweet
Details 詳細情報について
-
- CRID
- 1390295579819974528
-
- ISSN
- 18810225
- 18804535
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
-
- Abstract License Flag
- Disallowed