@inproceedings{li-etal-2026-security,
    title = "On the (In-)Security of the Shuffling Defense in the Transformer Secure Inference",
    author = "Li, Zhengyi  and
      Wang, Yakai  and
      Leng, Jingwen  and
      Yang, Kang  and
      Yu, Yu  and
      Gui, Jiaping  and
      Feng, Yu  and
      Liu, Ning  and
      Guo, Minyi",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/check-for-anonymous-pdfs/2026.acl-long.1341/",
    pages = "29084--29098",
    ISBN = "979-8-89176-390-6",
    abstract = "For Transformer models, cryptographically secure inference ensures that the client learns only the final output, while the server learns nothing about the client{'}s input. However, securely computing nonlinear layers remains a major efficiency bottleneck due to the substantial communication rounds and data transmission required. To address this issue, prior works reveal intermediate activations to the client, allowing nonlinear operations to be computed in plaintext. Although this approach significantly improves efficiency, exposing activations enables adversaries to extract model weights. To mitigate this risk, existing works employ a shuffling defense that reveals only randomly permuted activations to the client. In this work, we show that the shuffling defense is not as robust as previously claimed. We propose an attack that aligns differently shuffled activations to a common permutation and subsequently exploits them to extract model weights. Experiments on Pythia-70m and GPT-2 demonstrate that the proposed attack can align shuffled activations with mean squared errors ranging from $10^{-9}$ to $10^{-6}$. With a query cost of approximately {\$}1, the adversary can recover model weights with L1-norm differences ranging from $10^{-4}$ to $10^{-2}$ compared to the oracle weights."
}