Hyena: Balancing Packing, Reuse, and Rotations for Encrypted Inference
Sarabjeet Singh, Shreyas Singh, Sumanth Gudaparthi, Xiong Fan, Rajeev Balasubramonian
IEEE Symposium on Security and Privacy 2024 · Day 2 · Continental Ballroom 6
This talk introduces **Hyena**, a significant advancement towards achieving practical **privacy-preserving inference** using **Homomorphic Encryption (HE)**. Homomorphic Encryption is a powerful cryptographic tool that enables computations directly on encrypted data, unlocking the potential for sensitive applications such as medical imaging, genomics, and secure cloud-based compute outsourcing. However, its widespread adoption has been severely limited by substantial performance overheads, memory intensity, and inherent implementation complexity.
AI review
Hyena presents a truly clever, deeply technical solution to the long-standing performance bottleneck of rotations in Homomorphic Encryption for CNNs. By meticulously balancing packing, reuse, and delayed accumulation, it achieves near-zero permute-type rotations, making privacy-preserving inference practical with significant speed and memory efficiency gains. This isn't just theory; it's a foundational step towards secure AI.