Resource Consumption Forecasting and AI-Driven Service Scaling in 5G Environments.
In the 5th generation mobile networking, new types of services with high resource consumption and strict requirements in terms of quality of service (QoS) e.g. ultra low latency, high bandwidth, availability e.t.c, have emerged. The current thesis will exploit Machine Learning (ML) techniques so as to forecast future resource-related requests and proactively adapt the network resources in accordance to service resource demands (e.g. disk storage, memory size, cpu capacity, network bandwidth) in order to efficiently support service operation and ensure no Service-Level Agreement (SLA) violation.
Contact:
Lina Magoula
lina-magoula@di.uoa.gr
Network Anomaly Prediction and Forecasting in 5G Environments.
Securing the networks of tomorrow is set to be a challenging domain due to increased network heterogeneity, increased use of virtualisation technologies and distributed architectures. The current thesis will exploit well known Machine Learning (ML) techniques, targeting to to detect anomalous network traffic. To validate the effectiveness of the trained models, normal and anomalous network flows from a simulated environment using NS-3 tool have been collected.
Contact:
Lina Magoula
lina-magoula@di.uoa.gr
Simulating Anomalous Flows in Heterogeneous Networks
NS-3 has been developed to provide an open, extensible network simulation platform, for networking research and education. The current thesis is targeting to explore and implement different use case scenarios using NS-3 in order to simulate normal and anomalous network flows in heterogeneous networks (Wifi, LTE, 5G and Beyond).
Contact:
Lina Magoula
lina-magoula@di.uoa.gr
Predictive Contextual Profiling in 5G Environments using Distributed Machine Learning Techniques.
The advancement of technology and especially the introduction of the 5th generation mobile networking; introduces an ever-increasing amount of data flows. Managing and analysing this amount of data in a distributed manner is of major importance in order to proactively take measures to ensure a smooth and uninterrupted regardless-of-network-operation scale. This thesis will explore different Distributed Machine Learning Techniques towards network profile creation which will in turn provide improved accuracy, performance and larger input data size scaling.
Contact:
Nickolas Koursioumpas
nkoursioubas@di.uoa.gr
Distributed ML approaches for MEC-enabled Radio Resource Management for Beyond 5G Networks
The challenging environments and services of Beyond 5G (B5G) networks will rely ona great extent upon the application of AI-driven processes and Machine Learning(ML) algorithms for optimizing network resource management procedures. The target of this thesis focuses on the implementation and evaluation of a novelscheme for performing distributed inference and training of ML models in a B5Gnetwork.
Contact:
Sokratis Barmpounakis
sokbar@di.uoa.gr
Exploration of innovative AI-Driven Clustering Mechanisms towards Contextual Profiling in 5G Environments.
The 5th generation mobile networking is going to provide numerous advantages (e.g V2X Applications,Massive IoT Deployments) when compared to the existing 4G technologies . Additionally, there is a plethora of heterogeneous data sources that if managed properly, can provide considerable gains for the network. This thesis will explore different clustering mechanisms in order to collectively group this diverse information and create contextual profiles which will in turn provide useful insights with regards to network optimisation and management.
Contact:
Nickolas Koursioumpas
nkoursioubas@di.uoa.gr