AEGIS: Scaling Long-Sequence Homomorphic Encrypted Transformer Inference via Hybrid Parallelism on Multi-GPU Systems
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910102740533248 |
|---|---|
| author | Gong, Zhaoting Ran, Ran Yao, Fan Wen, Wujie |
| author_facet | Gong, Zhaoting Ran, Ran Yao, Fan Wen, Wujie |
| contents | Fully Homomorphic Encryption (FHE) enables privacy-preserving Transformer inference, but long-sequence encrypted Transformers quickly exceed single-GPU memory capacity because encoded weights are already large and encrypted activations grow rapidly with sequence length. Multi-GPU execution therefore becomes unavoidable, yet scaling remains challenging because communication is jointly induced by application-level aggregation and encryption-level RNS coupling. Existing approaches either synchronize between devices frequently or replicate encrypted tensors across devices, leading to excessive communication and latency.
We present AEGIS, an Application-Encryption Guided Inference System for scalable long-sequence encrypted Transformer inference on multi-GPU platforms. AEGIS derives device placement from ciphertext dependencies jointly induced by Transformer dataflow and CKKS polynomial coupling, co-locating modulus-coherent and token-coherent data so that communication is introduced only when application dependencies require it, while reordering polynomial operators to overlap the remaining collectives with computation.
On 2048-token inputs, AEGIS reduces inter-GPU communication by up to 57.9% in feed-forward networks and 81.3% in self-attention versus prior state-of-the-art designs. On four GPUs, it achieves up to 96.62% scaling efficiency, 3.86x end-to-end speedup, and 69.1% per-device memory reduction. These results establish coordinated application-encryption parallelism as a practical foundation for scalable homomorphic Transformer inference. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_03425 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AEGIS: Scaling Long-Sequence Homomorphic Encrypted Transformer Inference via Hybrid Parallelism on Multi-GPU Systems Gong, Zhaoting Ran, Ran Yao, Fan Wen, Wujie Cryptography and Security Artificial Intelligence Distributed, Parallel, and Cluster Computing Fully Homomorphic Encryption (FHE) enables privacy-preserving Transformer inference, but long-sequence encrypted Transformers quickly exceed single-GPU memory capacity because encoded weights are already large and encrypted activations grow rapidly with sequence length. Multi-GPU execution therefore becomes unavoidable, yet scaling remains challenging because communication is jointly induced by application-level aggregation and encryption-level RNS coupling. Existing approaches either synchronize between devices frequently or replicate encrypted tensors across devices, leading to excessive communication and latency. We present AEGIS, an Application-Encryption Guided Inference System for scalable long-sequence encrypted Transformer inference on multi-GPU platforms. AEGIS derives device placement from ciphertext dependencies jointly induced by Transformer dataflow and CKKS polynomial coupling, co-locating modulus-coherent and token-coherent data so that communication is introduced only when application dependencies require it, while reordering polynomial operators to overlap the remaining collectives with computation. On 2048-token inputs, AEGIS reduces inter-GPU communication by up to 57.9% in feed-forward networks and 81.3% in self-attention versus prior state-of-the-art designs. On four GPUs, it achieves up to 96.62% scaling efficiency, 3.86x end-to-end speedup, and 69.1% per-device memory reduction. These results establish coordinated application-encryption parallelism as a practical foundation for scalable homomorphic Transformer inference. |
| title | AEGIS: Scaling Long-Sequence Homomorphic Encrypted Transformer Inference via Hybrid Parallelism on Multi-GPU Systems |
| topic | Cryptography and Security Artificial Intelligence Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2604.03425 |