A multi-modal scalable high-performance homomorphic encryption system
Through software and hardware collaborative design, the memory scheduling and computing modules of the homomorphic encryption system are optimized, which solves the scheduling inefficiency problem of the multimodal homomorphic encryption system, achieves efficient ciphertext data processing and computing performance improvement, and adapts to application needs of different scales.
Patent Information
- Application Number
- CN202510026429.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-08
- Publication Date
- 2025-10-17
- Estimated Expiration
- 2045-01-08
AI Technical Summary
Existing homomorphic encryption systems suffer from inefficient scheduling in their hardware architecture design, especially in computing clusters of different sizes and multimodal homomorphic encryption application scenarios, which limits the reading and writing efficiency and computing efficiency of ciphertext data.
By adopting the hardware-software co-design method, we can achieve efficient management and parallel processing of multimodal homomorphic encryption algorithms by optimizing the scheduling strategy at the software level and introducing high-bandwidth memory, fast number theory transformation computing units and scalable computing modules at the hardware level.
It significantly improves the computing efficiency and performance of homomorphic encryption systems, solves the performance bottleneck caused by inefficient scheduling in traditional architectures, supports application scenarios of different scales, and improves the flexibility and compatibility of the system.