Congratulations to Dr. Xiali Hei and her post-doc fellow Md Imran Hossen for the NSF National AI Research Resource Pilot award

Congratulations to Dr. Xiali Hei and her post-doc fellow Md Imran Hossen! Their project “Investigating Security Issues in Instruction-Tuned Large Language Code Models” got the NSF National AI Research Resource Pilot award. In the first round, only 35 projects across USA were awarded.

In this project, the project aims to systematically uncover novel vulnerabilities and attack vectors introduced by instruction-tuned code LLMs, providing fundamental insights into their implications for cybersecurity and AI-incorporated software engineering workflows. The proposed work represents the first comprehensive investigation into the risks of code injection attacks against instruction-tuned code LLMs. Furthermore, we will develop a technique for identifying adversarial threats against instruction-tuned code LLMs, involving the detection of adversarially manipulated inputs and generated code before execution.
With the award, Dr. Hei’s group has the access to Delta at the National Center for Supercomputing Applications at the University of Illinois Urbana-Champaign and Frontera and Lonestar at the Texas Advanced Computing Center at The University of Texas at Austin for six months.

More details about the NAIRR Pilot program are available at the following link: https://nairrpilot.org/about.

A list of awarded projects is availble at the following link: https://nairrpilot.org/awarded-projects.