Research map / Large language models / Retrieval and knowledge
RAGFort: Dual-Path Defense Against Proprietary Knowledge Base Extraction in Retrieval-Augmented Generation
Qinfeng Li, Miao Pan, Ke Xiong, Ge Su, Zhiqiang Shen, Yan Liu, Sun Bing, Hao Peng, Xuhong Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add RAGFort dual-path defense | RAG system, Qwen-14B · HealthCareMagic RAG-Thief | CRR (%): 57.16 -> 27.96 | Improved |
| add RAGFort dual-path defense | RAG system · All datasets and models | relative mean CRR: 1x -> 0.51x | Improved |
| add re-ranking protection (inter-class) | RAG system · All datasets and models | relative mean CRR: 1x -> 0.91x | No clear effect |
| add summarization protection (intra-class) | RAG system · All datasets and models | relative mean CRR: 1x -> 0.87x | No clear effect |
| add contrastive reindexing to cascade-only defense (ablation) | RAGFort · All datasets and models | relative mean CRR: 0.83x -> 0.51x | Improved |
| add cascade generation to reindexing-only defense (ablation) | RAGFort · All datasets and models | relative mean CRR: 0.75x -> 0.51x | Improved |
| add contrastive reindexing only (ablation) | RAG system, Qwen-14B · Math QA RAG-Thief | CRR (%): 70.31 -> 54.61 | Improved |
| add cascade generation only (ablation) | RAG system, Qwen-14B · Math QA RAG-Thief | CRR (%): 70.31 -> 65.63 | No clear effect |
| add RAGFort (ablation) | RAG system, Qwen-14B · HealthCareMagic | answer accuracy (%): 68.16 -> 66.57 | Hurt |
| add RAGFort | RAG system, Qwen-14B · Math QA | answer accuracy (%): 85.31 -> 82.81 | Hurt |
| add RAGFort cascade with smaller draft model | RAG system, Qwen-14B · HealthCareMagic | FLOPs (T): 38.03 -> 19.40 | Improved |
| add RAGFort with same draft and reference model | RAG system, DeepSeek-R1-8B · HealthCareMagic | FLOPs (T): 20.37 -> 20.83 | No clear effect |
Closest papers by meaning
- On the Vulnerability of Applying Retrieval-Augmented Generation within Knowledge-Intensive Application Domains (ICML 2025 · similarity 0.88)
- SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG (NeurIPS 2025 · similarity 0.87)
- Follow My Instruction and Spill the Beans: Scalable Data Extraction from Retrieval-Augmented Generation Systems (ICLR 2025 · similarity 0.86)
- ReliabilityRAG: Effective and Provably Robust Defense for RAG-based Web-Search (NeurIPS 2025 · similarity 0.86)
- A Theory for Token-Level Harmonization in Retrieval-Augmented Generation (ICLR 2025 · similarity 0.85)
- From Complex to Atomic: Enhancing Augmented Generation via Knowledge-Aware Dual Rewriting and Reasoning (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
