Yinhao Hu (Huazhong University of Science and Technology & Zhongguancun Laboratory), Pengyu Ding (Huazhong University of Science and Technology & Zhongguancun Laboratory), Zhenpeng Lin (Independent Researcher), Dongliang Mu (Huazhong University of Science and Technology), Yuan Li (Zhongguancun Laboratory)

Despite extensive efforts to harden the Linux kernel—the foundation powering numerous widely-used distributions (e.g., Ubuntu, Debian, Fedora)—it continues to face persistent and sophisticated memory safety vulnerabilities. In this study, we introduce a novel systematic framework that decomposes kernel exploitation into three distinct phases from an attacker’s perspective. Through comprehensive analysis of 121 publicly documented exploits since 2015, we identify and categorize 64 recurrent attack vectors. Leveraging this structured approach, we perform an in-depth evaluation of 51 existing kernel defense mechanisms, clearly mapping their coverage, limitations, redundancies, and interdependencies. Our results reveal significant protection gaps: 23 attack vectors remain entirely unprotected, and 31 existing defenses are bypassable or obsolete. Additionally, we uncover notable discrepancies between theoretical effectiveness and practical deployment across popular downstream distributions, highlighting 4 underutilized hardening measures and misconfigurations in four major distributions. By illuminating these critical gaps and offering actionable insights, our work guides both kernel developers and security practitioners in enhancing defensive strategies and refining future security designs.

View More Papers

RTrace: Towards Better Visibility of Shared Library Execution

Huaifeng Zhang (Chalmers University of Technology), Ahmed Ali-Eldin (Chalmers University of Technology)

Read More

Mapping the Cloud: A Mixed-Methods Study of Cloud Security...

Sumair Ijaz Hashmi (CISPA Helmholtz Center for Information Security, Germany, Saarland University, Germany and Lahore University of Management Sciences (LUMS), Pakistan), Shafay Kashif (The University of Auckland, New Zealand and Lahore University of Management Sciences (LUMS), Pakistan), Lea Gröber (International Computer Science Institute (ICSI), USA and Lahore University of Management Sciences (LUMS), Pakistan), Katharina Krombholz…

Read More

Prompt Injection Attack to Tool Selection in LLM Agents

Jiawen Shi (Huazhong University of Science and Technology), Zenghui Yuan (Huazhong University of Science and Technology), Guiyao Tie (Huazhong University of Science and Technology), Pan Zhou (Huazhong University of Science and Technology), Neil Zhenqiang Gong (Duke University), Lichao Sun (Lehigh University)

Read More