Date of Award

4-18-2024

Publication Type

Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

Keywords

GPU Acceleration;Homomorphic Encryption;Post-Quantum Encryption

Supervisor

Mitra Mirhassani

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract

A new category of Fully Homomorphic Encryption (FHE) schemes, first presented in CHIMERA, offers methods of converting between different schemes to circumvent the disadvantages found inherently in each individually. The work done in PEGASUS introduced a similar agile Encryption framework, via an improved repacking algorithm that converts FHEW Ciphertexts into a CKKS Ciphertext. Using PEGASUS as a starting point, the goal was to speed up their novel repacking algorithm through the use of additional GPU computation. In doing so, the accelerated software created achieved a speedup of approximately 50x when comparing replaced portions of the algorithm now running in the GPU, and 25% reduction in the overall repacking algorithm runtime compared with the CPU only computation. This work focuses on the parts of the repacking algorithm that have thus far been unexplored by other GPU acceleration works, while leaving alone some other Homomorphic functions that have already been proven amenable to GPU acceleration. Also, the acceleration function is integrated in such a way that it builds alongside the current Open-Source PEGASUS framework, with minimal invasiveness, which increases potential for coupling this with complimentary works on acceleration of Homomorphic Encryption.

Download

Share

COinS