On predicting service-oriented network slices performances in 5G: A federated learning approach

Brik, Bouziane; Ksentini, Adlen

To achieve the vision of Zero Touch Management (ZSM) of network slices in 5G, it is
important to monitor and predict the performances of the running network slices, or their Key Performance Indicator (KPI). KPIs are usually monitored, but also, with the advance of Machine Learning (ML) techniques, predicted, in order to proactively react to
any service degradation of a running network slices. While network- and computation-oriented KPIs can be easily monitored and predicted, service-oriented KPIs are difficult to obtain due the privacy issue, as they disclose critical information on the performance of services. To tackle this issue, in this paper, we propose to use a new ML technique, known as Federated Learning (FL). It consists in keeping raw data where it is generated, while sending only users’ local trained models to the centralized entity for aggregation. Therefore, FL is adequate to predict slices’ service-oriented KPIs.

Systèmes de Communication
Eurecom Ref:
© 2020 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

PERMALINK : https://www.eurecom.fr/publication/6345