attack arXiv Apr 19, 2026 · 4w ago
Jiaqi Zhao, Fengwei Wang · Xidian University
Exploits subspace leakage in Euston's SVD-based transmission protocol to reconstruct private user inputs from secure transformer inference
Model Inversion Attack nlpvision
In the 47th IEEE Symposium on Security and Privacy (IEEE S&P 2026), Gao et al. proposed an efficient and user-friendly secure transformer inference framework, namely Euston. In Euston, a singular value decomposition-based matrix transmission protocol is designed to efficiently transmit input matrices, reducing communication bandwidth by approximately 2.8 times. In this manuscript, we show that this transmission protocol introduces subspace leakage of random masks, enabling the model owner to recover private samples easily. We further validate the effectiveness of the recovery attack through simple experiments on image and language datasets, highlighting a fundamental privacy risk of the protocol design.
transformer Xidian University
ML Security Papers