www.cerias.purdue.edu

Can/Will LLMs Learn to Reason?

Reasoning—the process of drawing conclusions from prior knowledge—is a hallmark of intelligence. Large language models, and more recently, large reasoning models have demonstrated impressive results ...
www.cerias.purdue.edu

EXPLAINABLE AI METHODS FOR ENHANCING AI-BASED NETWORK INTRUSION DETECTION SYSTEMS

In network security, the exponential growth of intrusions stimulates research toward developing advanced artificial intelligence (AI) techniques for intrusion detection systems (IDS). However, the ...
www.cerias.purdue.edu

Hanshen Xiao

Hanshen Xiao is an Assistant Professor in the Department of Computer Science. He received his Ph.D. degree in computer science from MIT and B.S. degree in Mathematics from Tsinghua University. Before ...
www.cerias.purdue.edu

Probabilistic Red-Teaming for Large Language and Vision-Language Models

As large language models (LLMs) and vision-language models (VLMs) grow more capable and widely deployed, they have also become increasingly susceptible to jailbreaks that bypass safety guardrails.
www.cerias.purdue.edu

Testing and detecting software upgrade failures in data-intensive distributed systems

In the current big data era, Internet services are often built on top of data-intensive distributed systems, such as distributed storage systems and distributed computation framework. Distributed ...
www.cerias.purdue.edu

Machine Unlearning

We analyze unlearning when a model must forget a designated subset while retaining utility on the rest, characterizing howsharpness‑oriented loss landscape via optimization interacts with ...
www.cerias.purdue.edu

Security and Privacy of Large Language Models (LLMs)

This project explores the security and privacy issues associated with Large Language Models (LLMs). Research in this area assesses the capability of LLMs to offer correct security advice and refute ...
www.cerias.purdue.edu

Security and Privacy in Augmented and Virtual Reality (AR/VR)

This project investigates emerging threats in mixed reality environments. The research includes analyzing user interface (UI) attacks in WebXR, exploring the extraction of sensitive information like ...
www.cerias.purdue.edu

Scalable Data Efficient Learning

We frame coreset selection (ie data subset selection) via posterior sampling so the subset’s induced loss landscape better matches the full‑data landscape, yielding robustness under label corruption ...
www.cerias.purdue.edu

Formalizing Enterprise Firewall Management with Informal and Elastic Specifications

Managing enterprise network firewalls is an ad-hoc process today where administrators must extract policy relevant to their enterprises from thousands of natural language vendor documents and tailor ...
www.cerias.purdue.edu

Active Learning using Loss Geometry

This work targets “atypical” points the model misclassifies with high confidence, shifting early rounds toward the right kind of hard examples. Aggregate tables show top average ranksover budgets and ...