Benchmarking Knowledge-Extraction Attack and Defense on Retrieval-Augmented Generation
Zhisheng Qi, Utkarsh Sahu, Li Ma, Haoyu Han, Ryan Rossi, Franck Dernoncourt, Mahantesh Halappanavar, Nesreen Ahmed, Yushun Dong, Yue Zhao, Yu Zhang, Yu Wang · 2026
Read Benchmarking Knowledge-Extraction Attack and Defense on Retrieval-Augmented Generation by Zhisheng Qi, Utkarsh Sahu, Li Ma, Haoyu Han, Ryan Rossi, Franck Dernoncourt, Mahantesh Halappanavar, Nesreen Ahmed, Yushun Dong, Yue Zhao, Yu Zhang, Yu Wang (2026) 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-2602-09319.txt.
Description
Retrieval-augmented generation pairs a language model with an outside store of documents, a setup now common in business chatbots, health assistants, and agent memory. A 2026 benchmark paper led by Zhisheng Qi, with eleven co-authors, warns that cleverly phrased questions can draw private store contents out through the model, threatening trade secrets and privacy. Until now, attacks and defenses had been tested in scattered ways that made fair comparison hard.
The paper gathers the field into one systematic benchmark covering many attack and defense strategies, several retrieval embeddings, open and closed generators, and both graph-based and plain indexing. All entries are judged under shared rules across datasets in several languages. By making results reproducible and directly comparable, the benchmark aims to give builders a firm basis for designing retrieval systems that keep their knowledge stores closed to extraction attempts.
Catalog record
| CREATOR | Zhisheng Qi, Utkarsh Sahu, Li Ma, Haoyu Han, Ryan Rossi, Franck Dernoncourt, Mahantesh Halappanavar, Nesreen Ahmed, Yushun Dong, Yue Zhao, Yu Zhang, Yu Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 83 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-09319