Saved in:
Bibliographic Details
Main Authors: Kale, Kaan, Mylonakis, Kyle, Roberts, Jay, Roy, Sidhartha
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.13758
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913848021221376
author Kale, Kaan
Mylonakis, Kyle
Roberts, Jay
Roy, Sidhartha
author_facet Kale, Kaan
Mylonakis, Kyle
Roberts, Jay
Roy, Sidhartha
contents In this work, we consider an inversion attack on the obfuscated input embeddings sent to a language model on a server, where the adversary has no access to the language model or the obfuscation mechanism and sees only the obfuscated embeddings along with the model's embedding table. We propose BeamClean, an inversion attack that jointly estimates the noise parameters and decodes token sequences by integrating a language-model prior. Against Laplacian and Gaussian obfuscation mechanisms, BeamClean always surpasses naive distance-based attacks. This work highlights the necessity for and robustness of more advanced learned, input-dependent methods.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13758
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BeamClean: Language Aware Embedding Reconstruction
Kale, Kaan
Mylonakis, Kyle
Roberts, Jay
Roy, Sidhartha
Cryptography and Security
In this work, we consider an inversion attack on the obfuscated input embeddings sent to a language model on a server, where the adversary has no access to the language model or the obfuscation mechanism and sees only the obfuscated embeddings along with the model's embedding table. We propose BeamClean, an inversion attack that jointly estimates the noise parameters and decodes token sequences by integrating a language-model prior. Against Laplacian and Gaussian obfuscation mechanisms, BeamClean always surpasses naive distance-based attacks. This work highlights the necessity for and robustness of more advanced learned, input-dependent methods.
title BeamClean: Language Aware Embedding Reconstruction
topic Cryptography and Security
url https://arxiv.org/abs/2505.13758