SUCCESS Lab
Posted on by guofei
We are excited to share that we have received a Fall 2025 Amazon Research Award!
This award will support our research on improving the security and trustworthiness of agentic AI systems, with a particular focus on understanding, benchmarking, and defending against prompt injection attacks.
Congratulations to the SUCCESS Lab team, and many thanks to Amazon for supporting our research!
Posted on by guofei
Today, Jianwei successfully defended his Ph.D. thesis, “Semantic-Aware Vulnerability Detection in Modern Web Applications.”
His research addresses critical challenges in web security by developing semantic-aware techniques for detecting vulnerabilities in complex, modern web applications. Notably, his work has identified numerous zero-day vulnerabilities (with many assigned CVEs) across a wide range of real-world web applications, demonstrating both the practical impact and technical depth of his contributions.
We are delighted to celebrate this well-deserved achievement and look forward to seeing the impact he will make in his next role at Palo Alto Networks.
Big congratulations, Dr. Huang! Wishing you continued success in this exciting new chapter.
Posted on by guofei
OpenClaw, a rapidly emerging open-source AI agent framework, is gaining tremendous momentum today. However, its powerful capabilities also introduce a broad attack surface. Without careful security design and deployment, OpenClaw-based agents could be compromised and abused as part of criminally controlled botnets.
In our new invited paper at ICCCN 2026, we discussed the security of OpenClaw. While it demonstrates the power of connecting LLM reasoning to real-world execution surfaces such as shell commands, filesystems, containers, browsers, and messaging platforms, this capability also greatly expands its attack surface. In this work, we present a systematic security taxonomy of 470 advisories filed against OpenClaw, organizing vulnerabilities by architectural layer and trust-violation type. Our analysis shows how weaknesses across gateway, execution policy, plugin/skill, sandbox, browser, and agent/prompt layers can compose into serious threats, including remote code execution paths, policy bypasses, prompt-injection-driven abuse, and supply-chain trust escalation. The study highlights an urgent lesson for the next generation of AI agent systems: powerful autonomy must be matched with unified, cross-layer security boundaries, or compromised agents may become attractive infrastructure for criminal abuse and botnet-scale control.
An extended version of the invited paper is available at arXiv:2603.27517. Congratulations on the nice work, Surada (our undergraduate researcher in the lab) and Yuxuan!
Posted on by guofei
Huancheng successfully defended his Ph.D. thesis, “Enabling Efficient, Effective, and Intelligent In-Network Monitoring on Programmable Switches,” today!
This milestone marks the culmination of years of hard work, creativity, and perseverance in advancing programmable switch-based network monitoring and security. We are very proud of his accomplishments and excited to see him begin the next chapter of his career at the Foundation for Science and AI Research (SAIR).
Big congratulations, Dr. Zhou! Wishing you continued success in your future research and professional journey!
Posted on by guofei
Our new paper on “Demystifying Progressive Web Application Permission Systems” is going to appear in DSN’26! Progressive Web Applications (PWAs) are web-based applications that bring native-app features such as offline access, notifications, and installability to the browser, making them an increasingly important part of today’s web ecosystem. In this work, we developed Permissioner to systematically analyze permission handling in PWAs and uncovered numerous security issues of inconsistency, incompleteness, and unclear boundaries in permission enforcement, with concrete cases of permission leakage, across browsers and platforms. Our findings have already led to fixes in Firefox and a Chromium patch, highlighting the urgent need for a unified and robust PWA permission model.
Congratulations, Mengxiao & Jianwei!
Posted on by guofei
Our new research CONSET (“Semantics Over Syntax: Uncovering Pre-Authentication 5G Baseband Vulnerabilities”) will appear in USENIX Security’26. This work leverages LLM and builds a new framework CONSET, which systematically extracts specification-level constraints and turns them into semantic violations for testing 5G UE implementations. On commercial smartphones, it confirms 6 device-level flaws through responsible disclosure, including 3 high-severity CVEs. These vulnerabilities affect 41 chipset models and over 466 commercially available smartphones.
This is joint work with researchers (Prof. Hongxin Hu and his students) at University of Buffalo. Congratulations on the amazing work, Qiqing!
Posted on by guofei
Our new research Heracles (“𝑶𝒏 𝒕𝒉𝒆 𝑺𝒆𝒄𝒖𝒓𝒊𝒕𝒚 𝑹𝒊𝒔𝒌𝒔 𝒐𝒇 𝑴𝒆𝒎𝒐𝒓𝒚 𝑨𝒅𝒂𝒑𝒕𝒂𝒕𝒊𝒐𝒏 𝒂𝒏𝒅 𝑨𝒖𝒈𝒎𝒆𝒏𝒕𝒂𝒕𝒊𝒐𝒏 𝒊𝒏 𝑫𝒂𝒕𝒂-𝒑𝒍𝒂𝒏𝒆 𝑫𝒐𝑺 𝑴𝒊𝒕𝒊𝒈𝒂𝒕𝒊𝒐𝒏”) will appear in NDSS’26. This work examines emerging security risks in modern programmable network infrastructures and highlights how design choices for efficiency can introduce new attack surfaces. We further propose Shield , a multi-layered mitigation scheme that successfully mitigates the Heracles attack while preserving line-rate performance and adaptive detection accuracy.
This is joint work with researchers (Prof. Min Suk Kang and his students) at KAIST. Congratulations on the amazing work, Hocheol! And thank you for spending time with us at the SUCCESS Lab as a visiting student in 2025!
Posted on by guofei
Our team, “All You Need Is A Fuzzing Brain”, led by Jeff, was one of seven finalists in DARPA’s Artificial Intelligence Cyber Challenge (AIxCC), placing 4th in the final round. During the competition, we developed a Cyber Reasoning System (CRS) that autonomously discovered 28 security vulnerabilities—including six previously unknown zero-days—in real-world open-source C and Java projects, and successfully patched 14 of them. A technical report is available here. The complete CRS is open source at here.
Big congratulations to the whole team, particularly Jeff and all TAMU students (Ze, Qingxiao, Jianwei, Matthew, etc)!
Posted on by guofei
Our DSN’15 paper “FloodGuard: A DoS Attack Prevention Extension in Software-Defined Networks” (by Haopei Wang, Lei Xu, Guofei Gu) received the Test of Time Award at this year’s DSN conference (2025 Annual IEEE/IFIP International Conference on Dependable Systems and Networks)! We are so honored and humbled to receive this recognition for our SDN security research in 2015! Congratulations again, Haopei & Lei!
Some background information about this award: “The Test-of-Time Award recognizes two outstanding papers published 10 years ago at DSN, in the DSN proceedings (research track, practical experience report or tool papers), that have had a sustained and important impact on the theory and/or practice of dependable systems and networks computing research. DSN has several areas under its umbrella and with two awards there are conditions to recognize more than one area. In exceptional situations (not enough nominations), the time frame for awards can be extended to 10-12 years, and only one paper can be awarded, in this order.”
Posted on by guofei
Our new paper on understanding the risks of dynamic content in the Alexa skill ecosystem is going to appear in ACM WiSec’25! Congratulations, Nathan!