Soutenance de thèse d'Imane Taïbi (ERMINE)

Vous êtes cordialement invités à venir assister à la soutenance de thèse d'Imane Taïbi, doctorante de l'équipe ERMINE, qui se tiendra le lundi 19 septembre à 10h00 en salle Métivier.

Sujet: Web-based data driven network monitoring : from performance estimation to anomaly detection

The goal of this thesis is to leverage passive measurements freely available in the browser and deep learning techniques to infer network performance and detect anomalies. We start by inferring the main properties of the underlying Network from web performance metrics based on passive measurements obtained from within the browser. We use machine learning to calibrate algorithms that allow such inference. By comparing deep learning algorithms to classical ML algorithms like Random Forest, we highlight the feasibility of the task but also its complexity hence the need for sophisticated deep learning algorithms such as convolutional neural networks (CNN). Then we study and examine the impact of web complexity on estimating two specific
metrics, delay and download bandwidth. Moreover, we propose an integrated framework to compare our approach with existing web-based monitoring solutions. Later, we propose an original network monitoring framework based on Bayesian Gaussian Mixture Models (BGMM) coupled with an algorithm to detect in real-time the occurrence of anomalies.