Xiangzhe Xu - Team leader
Xiangzhe Xu is a Ph.D. student focusing on code language model and program analysis. His research formulates program semantics via static and dynamic program analysis, and incorporates program domain knowledge in CodeLM via data augmentation and post-training. These techniques help CodeLM understand programs with limited symbol information.
Zian Su
Zian Su is a Ph.D student focusing on code language models and knowledge editing of LLMs. He has hands-on experience in fine-tuning large multi-modal models, and designing transformer architecture and pre-training objectives for code representation learning. His recent research involves cross-modal alignment of code models and retrieval augmented generation for low-level code understanding and reasoning.
Guangyu Shen
Guangyu Shen is a Ph.D. student focusing on AI security. He is the current lead of the Purdue team in the multi-round TrojAI competitions by IARPA and NIST. His team has won most of the competitions in the past 3.5 years. His research focuses on securing AI systems against a range of attacks, including backdoor and adversarial attacks.
Siyuan Cheng
Siyuan Cheng is a PhD student whose research expertise lies in the realm of trustworthy machine learning, with a specific focus on adversarial/backdoor attacks and defenses. He is one of weight-lifters in the TrojAI competitions. He was the team lead in the TDC2023 competition (for LLM backdoor scanning). His team achieved top rankings in the development phase.
Hanxi Guo
Hanxi Guo is a Ph.D. student focusing on AI security. He is the current lead of the Purdue team in the NIST GenAI competition. His team secures the second place among more than 15 teams from renowned research institutions and corporations in the blue teaming track, while also ranking in the top three in the red teaming track.
Xiaolong Jin
Xiaolong Jin is a PhD student in Computer Science at Purdue University, with a research focus on AI safety and efficiency, especially jailbreaking. He developed MULTIVERSE, an innovative jailbreaking tool that explores weakness in various contexts.
Lu Yan
Lu Yan is a PhD student, focusing on AI security with a particular interest in applying software security techniques to enhance the robustness of AI systems. She has experience in fine-tuning LLMs with QLoRA. She has developed ParaFuzz, a cutting-edge prompt paraphrasing/fuzzing technique to detect and remove backdoors.
Xuan Chen
Xuan Chen is a Ph.D. student focusing on developing RL-driven solutions to understand and mitigate AI vulnerabilities. She has developed RLbreaker, a novel technique that use reinforcement learning to construct jailbreaking prompts.
Jiasheng Jiang
Jiasheng Jiang is a PhD student whose expertise is on software vulnerability detection. His vulnerability detector has found hundreds of vulnerabilities in Linux kernel, Openssl, Ffmepg, and Apache https.
Xiangyu Zhang - Faculty advisor
Xiangyu Zhang is an expert in AI security, program analysis, vulnerability detection, and malware analysis. He has led successful projects with DARPA, IARPA, and ONR, notably winning rounds in IARPA TrojAI competitions. His research group of 20 PhD students and 2 post-docs has developed tools for detecting AI backdoors and software vulnerabilities.
Zhuo Zhang - Faculty advisor
Zhuo Zhang is a renowned hacker with significant real-world impact, including preventing a $2 million loss. He led teams to victory in DEFCON CTF 2020 and Paradigm CTF 2023. His award-winning research in software and malware analysis earned him the 2024 ACM SIGSAC Distinguished Dissertation Award.
Chengpeng Wang - Faculty advisor
Chengpeng Wang specializes in static analysis to improve software reliability and performance. His bug detection methods are used in industry, and his current work integrates neurosymbolic static analysis techniques like LLMDFA and LLMSAN. These approaches use LLMs and symbolic analysis to suppress hallucinations and allow customization without compilation.
