Guanlong Wu (Southern University of Science and Technology), Taojie Wang (Southern University of Science and Technology), Yao Zhang (ByteDance Inc.), Zheng Zhang (Southern University of Science and Technolog), Jianyu Niu (Southern University of Science and Technology), Ye Wu (ByteDance Inc.), Yinqian Zhang (SUSTech)

The emergence of large language models (LLMs) has enabled a wide range of applications, including code generation, chatbots, and AI agents. However, deploying these applications faces substantial challenges in terms of cost and efficiency. One notable optimization to address these challenges is semantic caching, which reuses query-response pairs across users based on semantic similarity. This mechanism has gained significant traction in both academia and industry and has been integrated into the LLM serving infrastructure of cloud providers such as Azure, AWS, and Alibaba.

This paper is the first to show that semantic caching is vulnerable to cache poisoning attacks, where an attacker injects crafted cache entries to cause others to receive attacker-defined responses. We demonstrate the semantic cache poisoning attack in diverse scenarios and confirm its practicality across all three major public clouds. Building on the attack, we evaluate existing adversarial prompting defenses and find they are ineffective against semantic cache poisoning, leading us to propose a new defense mechanism that demonstrates improved protection compared to existing approaches, though complete mitigation remains challenging. Our study reveals that cache poisoning, a long-standing security concern, has re-emerged in LLM systems. While our analysis focuses on semantic cache, the underlying risks may extend to other types of caching mechanisms used in LLM systems.

View More Papers

“I found the text to be encouraging” – Evaluating...

Rozalina Doneva (Karlsruhe Institute of Technology (KIT)), Anne Hennig (Karlsruhe Institute of Technology (KIT)), Peter Mayer (University of Southern Denmark (SDU))

Read More

WCDCAnalyzer: Scalable Security Analysis of Wi-Fi Certified Device Connectivity...

Zilin Shen (Purdue University), Imtiaz Karim (The University of Texas at Dallas), Elisa Bertino (Purdue University)

Read More

Icarus: Achieving Performant Asynchronous BFT with Only Optimistic Paths

Xiaohai Dai (Huazhong University of Science and Technology), Yiming Yu (Huazhong University of Science and Technology), Sisi Duan (Tsinghua University), Rui Hao (Wuhan University of Technology), Jiang Xiao (Huazhong University of Science and Technology), Hai Jin (Huazhong University of Science and Technology)

Read More