Name of the participant: Patrick Kalmbach
Description of the IT research project: SELMA investigates to what extent AI techniques can be used to learn comprehensible and meaningful models for the behaviour of individual distributed applications or entire networks automatically and to use them for network control. For example, such a model can automatically break down communicating nodes into structural and/or functional groups and characterize the communication between these groups. This information can be valuable for the recognition of applications. The model also generates a compact representation that can be used by downstream systems, for example for detecting malware. By means of new techniques for network control, this can be automatically applied to the network.
The figure shows the schematic structure of SELMA. SELMA therefore touches on the modules Model of Communication, Application Recognition and Network Control. Different techniques of network monitoring, such as DPI, the reading of information from network control devices or dedicated network monitoring systems create the data basis. With the help of AI a model of the communication is to be extracted from this data basis. The AI prepares the data and generates information that supports the recognition of applications and can be used directly for network control. The application recognition shall also use AI to classify encrypted traffic. The information obtained from the model and the application recognition finally flows together in the network control. In order to be able to automatically derive control signals for the network, the use of AI is also required here.
Software Campus partners: TU Munich, Rohde & Schwarz
Implementation period: 01.01.2020 – 31.03.2021