Rethinking the Privacy of Text Embeddings: A Reproducibility Study of "Text Embeddings Reveal (Almost) As Much As Text"

• URL: http://arxiv.org/abs/2507.07700v1
• Date: Thu, 10 Jul 2025 12:27:03 GMT
• Title: Rethinking the Privacy of Text Embeddings: A Reproducibility Study of "Text Embeddings Reveal (Almost) As Much As Text"
• Authors: Dominykas Seputis, Yongkang Li, Karsten Langerak, Serghei Mihailov,
• Abstract summary: We reproduce the Vec2Text framework and evaluate it from two perspectives.<n>We successfully replicate the original key results in both in-domain and out-of-domain settings.<n>We identify key limitations, including its sensitivity to input sequence length.
• Score: 0.14061979259370275
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Text embeddings are fundamental to many natural language processing (NLP) tasks, extensively applied in domains such as recommendation systems and information retrieval (IR). Traditionally, transmitting embeddings instead of raw text has been seen as privacy-preserving. However, recent methods such as Vec2Text challenge this assumption by demonstrating that controlled decoding can successfully reconstruct original texts from black-box embeddings. The unexpectedly strong results reported by Vec2Text motivated us to conduct further verification, particularly considering the typically non-intuitive and opaque structure of high-dimensional embedding spaces. In this work, we reproduce the Vec2Text framework and evaluate it from two perspectives: (1) validating the original claims, and (2) extending the study through targeted experiments. First, we successfully replicate the original key results in both in-domain and out-of-domain settings, with only minor discrepancies arising due to missing artifacts, such as model checkpoints and dataset splits. Furthermore, we extend the study by conducting a parameter sensitivity analysis, evaluating the feasibility of reconstructing sensitive inputs (e.g., passwords), and exploring embedding quantization as a lightweight privacy defense. Our results show that Vec2Text is effective under ideal conditions, capable of reconstructing even password-like sequences that lack clear semantics. However, we identify key limitations, including its sensitivity to input sequence length. We also find that Gaussian noise and quantization techniques can mitigate the privacy risks posed by Vec2Text, with quantization offering a simpler and more widely applicable solution. Our findings emphasize the need for caution in using text embeddings and highlight the importance of further research into robust defense mechanisms for NLP systems.

Related papers

• TextSleuth: Towards Explainable Tampered Text Detection [49.88698441048043]
  We propose to explain the basis of tampered text detection with natural language via large multimodal models.<n>To fill the data gap for this task, we propose a large-scale, comprehensive dataset, ETTD.<n>Elaborate queries are introduced to generate high-quality anomaly descriptions with GPT4o.<n>To automatically filter out low-quality annotations, we also propose to prompt GPT4o to recognize tampered texts.
  arXiv  Detail & Related papers  (2024-12-19T13:10:03Z)
• Leveraging Structure Knowledge and Deep Models for the Detection of Abnormal Handwritten Text [19.05500901000957]
  We propose a two-stage detection algorithm that combines structure knowledge and deep models for handwritten text. A shape regression network trained by a novel semi-supervised contrast training strategy is introduced and the positional relationship between the characters is fully employed. Experiments on two handwritten text datasets show that the proposed method can greatly improve the detection performance.
  arXiv  Detail & Related papers  (2024-10-15T14:57:10Z)
• Understanding and Mitigating the Threat of Vec2Text to Dense Retrieval Systems [28.175920880194223]
  We investigate factors related to embedding models that may impact text recoverability via Vec2Text. We propose a simple embedding transformation fix that guarantees equal ranking effectiveness while mitigating the recoverability risk.
  arXiv  Detail & Related papers  (2024-02-20T07:49:30Z)
• SwinTextSpotter v2: Towards Better Synergy for Scene Text Spotting [121.44909266398194]
  We propose a new end-to-end scene text spotting framework termed SwinTextSpotter v2.<n>We enhance the relationship between two tasks using novel Recognition Conversion and Recognition Alignment modules.<n>SwinTextSpotter v2 achieved state-of-the-art performance on various multilingual (English, Chinese, and Vietnamese) benchmarks.
  arXiv  Detail & Related papers  (2024-01-15T12:33:00Z)
• On the Possibilities of AI-Generated Text Detection [76.55825911221434]
  We argue that as machine-generated text approximates human-like quality, the sample size needed for detection bounds increases. We test various state-of-the-art text generators, including GPT-2, GPT-3.5-Turbo, Llama, Llama-2-13B-Chat-HF, and Llama-2-70B-Chat-HF, against detectors, including oBERTa-Large/Base-Detector, GPTZero.
  arXiv  Detail & Related papers  (2023-04-10T17:47:39Z)
• SwinTextSpotter: Scene Text Spotting via Better Synergy between Text Detection and Text Recognition [73.61592015908353]
  We propose a new end-to-end scene text spotting framework termed SwinTextSpotter. Using a transformer with dynamic head as the detector, we unify the two tasks with a novel Recognition Conversion mechanism. The design results in a concise framework that requires neither additional rectification module nor character-level annotation.
  arXiv  Detail & Related papers  (2022-03-19T01:14:42Z)
• Evaluating the Construct Validity of Text Embeddings with Application to Survey Questions [5.343406649012618]
  We propose the use of the classic construct validity framework to evaluate the validity of text embeddings. We show how this framework can be adapted to the opaque and high-dimensional nature of text embeddings. We show that embeddings can be used to predict respondent's answers to completely new survey questions.
  arXiv  Detail & Related papers  (2022-02-18T12:35:46Z)
• CORE-Text: Improving Scene Text Detection with Contrastive Relational Reasoning [65.57338873921168]
  Localizing text instances in natural scenes is regarded as a fundamental challenge in computer vision. In this work, we quantitatively analyze the sub-text problem and present a simple yet effective design, COntrastive RElation (CORE) module. We integrate the CORE module into a two-stage text detector of Mask R-CNN and devise our text detector CORE-Text.
  arXiv  Detail & Related papers  (2021-12-14T16:22:25Z)
• On Exploring and Improving Robustness of Scene Text Detection Models [20.15225372544634]
  We evaluate scene text detection models ICDAR2015-C (IC15-C) and CTW1500-C (CTW-C) We perform a robustness analysis of six key components: pre-training data, backbone, feature fusion module, multi-scale predictions, representation of text instances and loss function. We present a simple yet effective data-based method to destroy the smoothness of text regions by merging background and foreground.
  arXiv  Detail & Related papers  (2021-10-12T02:36:48Z)
• ABCNet v2: Adaptive Bezier-Curve Network for Real-time End-to-end Text Spotting [108.93803186429017]
  End-to-end text-spotting aims to integrate detection and recognition in a unified framework. Here, we tackle end-to-end text spotting by presenting Adaptive Bezier Curve Network v2 (ABCNet v2) Our main contributions are four-fold: 1) For the first time, we adaptively fit arbitrarily-shaped text by a parameterized Bezier curve, which, compared with segmentation-based methods, can not only provide structured output but also controllable representation. Comprehensive experiments conducted on various bilingual (English and Chinese) benchmark datasets demonstrate that ABCNet v2 can achieve state-of-the
  arXiv  Detail & Related papers  (2021-05-08T07:46:55Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.