Pipelined Key Switching Accelerator Architecture for CKKS-Based Fully Homomorphic Encryption.

Area of Science:

• Cryptography
• Computer Architecture
• Data Security

Background:

• Big data and cloud computing raise privacy concerns.
• Fully homomorphic encryption (FHE) allows computation on encrypted data but is computationally expensive.
• Optimization and acceleration are crucial for practical FHE adoption.

Purpose of the Study:

• To introduce an efficient hardware architecture, the KeySwitch module, for accelerating FHE.
• To address the computational and memory challenges of homomorphic evaluations.
• To enhance the practical applicability of FHE in privacy-preserving computations.

Main Methods:

• Designed a highly efficient and pipelined hardware architecture (KeySwitch module).
• Leveraged an area-efficient number-theoretic transform design.
• Incorporated fine-grained pipelining, optimized on-chip resource usage, and high-throughput implementation.

Main Results:

• Achieved a 1.6x improvement in data throughput on an FPGA platform.
• Demonstrated more efficient hardware resource utilization compared to prior work.
• Validated the effectiveness of the KeySwitch module in accelerating key switching operations.

Conclusions:

• The KeySwitch module offers significant acceleration for homomorphic computations.
• This hardware accelerator enhances efficiency for privacy-preserving applications.
• The work promotes the practical adoption of FHE by improving performance and resource utilization.