Kaihua Wang (Tsinghua University), Jianjun Chen (Tsinghua University), Pinji Chen (Tsinghua University), Jianwei Zhuge (Tsinghua University), Jiaju Bai (Beihang University), Haixin Duan (Tsinghua University)

QUIC is a modern transport protocol increasingly adopted by major platforms and services, making its security and correctness critically important. However, the complexity of QUIC specification and implementations introduces opportunities for subtle and dangerous logic flaws. Existing QUIC testing tools primarily focus on memory-related vulnerabilities and are ill-equipped to detect logical vulnerabilities. Therefore, the discovery of logical vulnerabilities is currently still highly dependent on manual auditing.

In this paper, we introduce MerCuriuzz, a novel black-box fuzzing framework designed to automatically uncover logical vulnerabilities in QUIC implementations. We evaluated MerCuriuzz against 16 widely used QUIC implementations and discovered 14 previously unknown logical vulnerabilities affecting popular implementations such as quiche, xquic, and aioquic. Those vulnerabilities can pose severe security risks, enabling attackers to exhaust server resources, crash services, or deny legitimate users access to the server. We categorize those vulnerabilities into six categories and propose mitigation strategies. We also responsibly disclosed our findings to the affected vendors, and 11 of them were confirmed and rewarded by the vendors, such as Cloudflare and Alibaba Cloud.

View More Papers

Entente: Cross-silo Intrusion Detection on Network Log Graphs with...

Jiacen Xu (Microsoft), Chenang Li (University of California, Irvine), Yu Zheng (University of California, Irvine), Zhou Li (University of California, Irvine)

Read More

TBTrackerX: Fantastic Trigger Bots and Where to Find Malicious...

Mohammad Majid Akhtar (University of New South Wales), Rahat Masood (University of New South Wales), Muhammad Ikram (Macquarie University), Salil S. Kanhere (University of New South Wales)

Read More

CAT: Can Trust be Predicted with Context-Awareness in Dynamic...

Jie Wang (Xidian University), Zheng Yan (Xidian University), Jiahe Lan (Xidian University), Xuyan Li (Xidian University), Elisa Bertino (Purdue University)

Read More