Marija FURDEKĀ  (Chalmers University of Technology)

Monitoring optical network security with machine learning

Optical networks are critical infrastructure vulnerable to a range of physical layer attacks that can degrade a multitude of services in the upper network layers. Despite recent progress, a framework for cognitive security management is not established yet. In this talk, we will discuss the challenges related to attack prevention, detection, and recovery. The focus will be placed on supervised, unsupervised and semi-supervised learning techniques that have promising performance in detecting known as well as novel attack types. The main challenges related to accuracy, scalability, interpretability and practical deployment of ML-based security diagnostics procedures will be reviewed, and ways of addressing these challenges will be examined.