"CyLab Faculty, Students to Present at NDSS Symposium 2024"

Faculty and students from CyLab, Carnegie Mellon University's security and privacy research institute, will present on various topics at the 31st Annual Network and Distributed System Security (NDSS) Symposium. CyLab has compiled a list of papers co-authored by its members that will be presented at the event. One of the papers is titled "Group-based Robustness: A General Framework for Customized Robustness in the Real World." Machine Learning (ML) models have been found to be vulnerable to evasion attacks that perturb model inputs and cause misclassifications. In this paper, researchers identify real-world scenarios in which existing attacks cannot accurately assess the true threat. They found that traditional metrics for measuring targeted and untargeted robustness do not accurately reflect a model's capacity to withstand attacks from one set of source classes to another set of target classes. The researchers defined a new metric called group-based robustness, which complements existing metrics and better evaluates model performance in specific attack scenarios. This article continues to discuss papers co-authored by CyLab faculty and students that will be presented at the NDSS Symposium.

CyLab reports "CyLab Faculty, Students to Present at NDSS Symposium 2024"

Submitted by grigby1