SUPR
Machine learning for adversarial environments
Dnr:

NAISS 2023/5-465

Type:

NAISS Medium Compute

Principal Investigator:

György Dán

Affiliation:

Kungliga Tekniska högskolan

Start Date:

2023-12-01

End Date:

2024-12-01

Primary Classification:

10202: Information Systems (Social aspects to be 50804)

Allocation

Abstract

The project investigates robust machine learning algorithms in adversarial environments. Motivated by the well known vulnerability of deep neural networks to norm-bounded perturbation attacks and physically realizable attacks, the objective of the project is to develop algorithms and methods for detecting attacks combining Bayesian inference with sequential detection, and to use these algorithms and methods for developing effective methods and frameworks for incident response automation.