Somewhat Homomorphic Encryption from Linear Homomorphism and Sparse LPN
summary
Construction claims somewhat homomorphic encryption from sparse LPN plus a standard assumption that yields linearly homomorphic encryption (e.g., DDH or DCR), avoiding lattices and pairings. Ciphertexts are matrices as ciphertexts with homomorphic addition = matrix addition and multiplication = matrix multiplication; decrypting is a cheap linear function. Scheme supports o(log λ) multiplications followed by poly(λ) additions over Fq, needs a large evaluation key (stored once) but the construction and security proofs appear to work (EUROCRYPT 2025 paper).
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:01Homomorphic encryption spectrum
- 6:13Motivation for new HE assumptions
- 8:23Sparse LPN overview
- 14:26Homomorphic ops with sparse LPN
- 27:19Non-compact ciphertexts challenge
- 31:36Sparse LPN with linear HE
- 37:12Final construction and key eval
- 40:30Efficiency and assumptions open questions
- 43:19Approaches to achieve fully homomorphic encryption
- 51:00Q&A








