Yinan Zhong (Zhejiang University), Qianhao Miao (Zhejiang University), Yanjiao Chen (Zhejiang University), Jiangyi Deng (Zhejiang University), Yushi Cheng (Zhejiang University), Wenyuan Xu (Zhejiang University)

Large Language Models (LLMs) have been integrated into many applications (e.g., web agents) to perform more sophisticated tasks. However, LLM-empowered applications are vulnerable to Indirect Prompt Injection (IPI) attacks, where instructions are injected via untrustworthy external data sources. This paper presents Rennervate, a defense framework to detect and prevent IPI attacks. Rennervate leverages attention features to detect the covert injection at a fine-grained token level, enabling precise sanitization that neutralizes IPI attacks while maintaining LLM functionalities. Specifically, the token-level detector is materialized with a 2-step attentive pooling mechanism, which aggregates attention heads and response tokens for IPI detection and sanitization. Moreover, we establish a fine-grained IPI dataset, FIPI, to be open-sourced to support further research. Extensive experiments verify that Rennervate outperforms 15 commercial and academic IPI defense methods, achieving high precision on 5 LLMs and 6 datasets. We also demonstrate that Rennervate is transferable to unseen attacks and robust against adaptive adversaries.

View More Papers

The Compromised Satellite Peripheral Dilemma

Rachel McAmis (MIT Lincoln Laboratory and University of Washington), Connor Willison (MIT Lincoln Laboratory), Richard Skowyra (MIT Lincoln Laboratory), Samuel Mergendahl (MIT Lincoln Laboratory)

Read More

Scalable Off-Chain Auctions

Mohsen Minaei (Visa Research), Ranjit Kumaresan (Visa Research), Andrew Beams (Visa Research), Pedro Moreno-Sanchez (IMDEA Software Institute, MPI-SP), Yibin Yang (Georgia Institute of Technology), Srinivasan Raghuraman (Visa Research and MIT), Panagiotis Chatzigiannis (Visa Research), Mahdi Zamani (Visa Research), Duc V. Le (Visa Research)

Read More

Mobius: Enabling Byzantine-Resilient Single Secret Leader Election with Uniquely...

Hanyue Dou (Institute of Software, Chinese Academy of Sciences; the School of Computer Science and Technology, University of Chinese Academy of Sciences), Peifang Ni (Institute of Software, Chinese Academy of Sciences; Zhongguancun Laboratory), Yingzi Gao (Shandong University), Jing Xu (Institute of Software, Chinese Academy of Sciences; Zhongguancun Laboratory)

Read More