Configurable Encryption and Decryption Architectures for CKKS-Based Homomorphic Encryption.

Area of Science:

• Cryptography
• Hardware Architecture
• Edge Computing

Background:

• Growing concerns regarding data privacy and security with increased edge device connectivity to the cloud.
• Homomorphic Encryption (HE) offers a solution for privacy-preserving computations on encrypted data.
• Limited optimization efforts for HE encryption/decryption (ENC-DEC) operations on edge devices.

Purpose of the Study:

• To propose efficient hardware architectures for CKKS-based ENC-DEC accelerators.
• To facilitate and optimize client-side computations using homomorphic encryption.
• To address the performance bottleneck of ENC-DEC operations at the edge.

Main Methods:

• Design of configurable hardware architectures for CKKS-based ENC-DEC.
• Support for a wide range of polynomial sizes and multiplicative depths (up to 30 levels).
• Implementation and evaluation on the Xilinx XCU250 FPGA platform with a 128-bit security guarantee.

Main Results:

• Achieved an average encryption time 23.7x faster compared to the SEAL HE library.
• Demonstrated efficient hardware utilization for cryptographic algorithms.
• Configurable architectures supporting various parameters for CKKS HE.

Conclusions:

• The proposed hardware architectures significantly accelerate client-side ENC-DEC operations.
• These designs enhance data privacy and computational capabilities for edge devices.
• Potential to accelerate cryptographic processes in the post-quantum era, improving edge security.