ETD

• Acceleration
• AI
• ARM SVE
• Homomorphic Encryption
• HPC
• Microsoft SEAL

The thesis focused on optimizing the multiplication used by the CKKS scheme in Homomorphic Encryption computations on HPC and dedicated hardware platforms. After analyzing the mathematical structure and performance limitations of existing HE libraries (Microsoft SEAL), the work introduced a vectorized implementation based on the ARM Scalable Vector Extension (SVE) to accelerate said multiplication within SEAL while maintaining full computational correctness. Benchmark comparison between standard SEAL, SEAL+SVE, and SEAL+Intel HEXL were carried out to quantify improvements in speed and scalability. In parallel, for a more complete overview of all the possibilities, a hardware acceleration path inspired by the SEAL architecture was designed. Overall, the research demonstrates that hardware-aware optimization of homomorphic computation can reduce execution times while maintaining numerical precision, paving the way for practical adoption of HE in high-performance computing applications.