ALGEN: Few-shot Inversion Attacks on Textual Embeddings using Alignment and Generation
Yiyi Chen, Qiongkai Xu, Johannes Bjerva · 2025
Read ALGEN: Few-shot Inversion Attacks on Textual Embeddings using Alignment and Generation by Yiyi Chen, Qiongkai Xu, Johannes Bjerva (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-11308.txt.
Description
Yiyi Chen and co-authors present ALGEN in a 2025 paper, a few-shot attack that rebuilds private text from stored embeddings. Text is now often kept as number vectors, yet those vectors can leak the words they came from.
Past attacks needed millions of sentences to train. ALGEN aligns the target vectors to an attack space and uses a text generator to rebuild the source. One sample can partly work, and about a thousand samples reach peak results across many encoders. The attack moves across topics and tongues, and no tested defense stopped it. Privacy teams get a stark warning that embeddings are not safe storage.
Catalog record
| CREATOR | Yiyi Chen, Qiongkai Xu, Johannes Bjerva |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-11308