Abstract:
This talk will give an overview about recent research results on reframing known challenges of computer networks in a framework usable by neural network. The main concept is based on modeling network architectures and their configuration as annotated graphs. Using recent advances from the deep learning community on graph processing, those graph representations can then be used in conjunction with Graph Neural Networks for reasoning about computer network architectures. After an overview about various problems which have been tackled with this approach, the use-case of verification of MPLS networks will be presented. Our approach, called DeepMPLS, focuses on automated verification of the policy-compliance of network configurations and repair of broken ones.
Short bio:
Fabien Geyer is a researcher at Airbus Central Research & Technology in Munich (Germany), working working on methods for network analytics, network performances and architectures. He received the master of engineering in telecommunications from Telecom Bretagne, Brest, France in 2011 and the Ph.D. degree in computer science from Technical University of Munich (TUM), Munich, Germany in 2015. His research interests include formal methods for the performance evaluation and modeling of network architectures and protocols.
