Understanding Attacks On AI
12-29, 18:30–19:30 (Europe/Berlin), HDMI (room 3)
Language: English

The field of AI security, or more specifically ML security, is only just developing. Although a growing number of attack types on ML training data, training, models, and deployments have been discovered, there is no agreement on a clear taxonomy yet. In this workshop we will look at different ML attack types and try to come up with a systematization and a mapping to cyber/infosec attack types and tools.


The field of AI security, or more specifically ML security, is only just developing. Although a growing number of attack types on ML training data, training, models, and deployments have been discovered, there is no agreement on a clear taxonomy yet. In this workshop we will look at different ML attack types and try to come up with a systematization and a mapping to cyber/infosec attack types and tools.