活动预告｜CodeWisdom 可信AI系统系列学术报告第13期

内容简介

Semantic caching has emerged as a pivotal technique for scaling LLM applications, widely adopted by major providers including AWS and Microsoft. By utilizing semantic embedding vectors as cache keys, this mechanism effectively minimizes latency and redundant computation for semantically similar queries. In this work, we conceptualize semantic cache keys as a form of fuzzy hashes. We demonstrate that the locality required to maximize cache hit rates fundamentally conflicts with the cryptographic avalanche effect necessary for collision resistance. Our conceptual analysis formalizes this inherent trade-off between performance (locality) and security (collision resilience), revealing that semantic caching is naturally vulnerable to key collision attacks.

While prior research has focused on side-channel and privacy risks, we present the first systematic study of integrity risks arising from cache collisions. We introduce CacheAttack, an automated framework for launching black-box collision attacks. We evaluate CacheAttack in security-critical tasks and agentic workflows. It achieves a hit rate of 86\% in LLM response hijacking and can induce malicious behaviors in LLM agent, while preserving strong transferability across different embedding models. A case study on a financial agent further illustrates the real-world impact of these vulnerabilities. Finally, we discuss mitigation strategies.

Zhixiang Zhang(张芝翔) recently received his B.Eng. in Software Engineering from Fudan University. In Fall 2026, he will join the Hong Kong University of Science and Technology (HKUST) as a PhD student, advised by Prof. Dongdong She. His research primarily focuses on AI security, specifically the robustness and privacy of Large Language Models (LLMs) and Trustworthy AI. His recent work on LLM semantic caching and agent security has been accepted at top-tier conferences including ICML, ASE, and ISSTA. Beyond his research, he actively engages with the academic community by serving as a reviewer and sub-reviewer for conferences like ACL, ICML, ISSTA, ASE, CCS and so on.

时间安排

时间：2026年7月22日10:30

地点：复旦大学江湾校区交叉二号楼A2003

腾讯会议：474-813-999  密码：169539