Using Level Homomorphic Encryption in a Client-Server Setting to Evaluate Artificial Neural Networks Against Encrypted Inputs
Patent Information
- Application Number
- JP2024533302
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-12-14
- Filing Date
- 2022-04-08
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-04-08
AI Technical Summary
Existing Level Homomorphic Encryption (LHE) schemes are inefficient for deep Artificial Neural Networks (ANNs) due to high noise thresholds requiring large ciphertexts, and they either reveal private information or are computationally inefficient, and they only support integer operations, not real-valued data.
A computer-implemented method using randomness-based blinding and rescaling techniques to reduce ciphertext noise and scale factors, enabling efficient computations on encrypted inputs without revealing intermediate results, supporting deep ANNs and real-valued data through interactive bootstrapping and secure multi-party computation.
Enables efficient computation on encrypted inputs with reduced noise and scale factors, allowing deep ANNs and real-valued operations without leaking private information, improving computational efficiency and privacy in client-server settings.
Smart Images

Figure 2024544374000001_ABST
Abstract
Citation Information
Patent Citations
A method and apparatus for securing encryption keys in insecure computer environments, applicable to the security and management of virtualization and cloud computing.
JP2015503280A
Commission of processing operation using homomorphism code
JP2019191575A
Encryption method and device using homomorphic encryption
JP2021177239A
Method and apparatus for processing ciphertext based on homomorphic encryption
JP2021179603A
Fully homomorphic encryption method based on a bootstrappable encryption scheme, computer program and apparatus
US20110110525A1
