PPGI PhD thesis defese (Friday 21st at 9am )
Dear all,
We cordially invite you to attend the PhD thesis defense of Bruno Missi Xavier, titled "Crossing Domains for Accuracy: In-Network Stacking of Machine Learning Classifiers."
Date: Friday, 21st
Time: 9:00 AM
The defense will take place in a hybrid format at room 34 of CT13 and also virtually at https://meet.google.com/pga-nimx-nvm.
Abstract: This research marks a significant shift from traditional traffic classification techniques, which are typically end-host based, by embedding ML directly into the network.
Key contributions of the thesis include:
MAP4: Demonstrates the deployment of ML models (such as Decision Trees and Random Forests) within the data plane using the P4 language, enabling accurate flow classification at line rate.
In-Network Concept Drift: Introduces a method to detect changes in network traffic distribution by implementing Exponentially Weighted Moving Average (EWMA) to address the limitations of P4.
Eagle Framework: Designed as a defense mechanism against cyber threats by utilizing Open Radio Access Network (O-RAN) to collect measurements from the air interface (PHY and MAC layers) for early detection and mitigation of malicious flows.
Cross-Domain AI: Integrates multiple layers (RAN and programmable data planes) to create an in-network stacking of ML classifiers under a multi-view learning approach, thereby enhancing the overall accuracy of traffic classification systems.
Members of the jury:
Prof. Magnos Martinello - UFES - supervisor
Prof. Marco Ruffini - Trinity College Dublin (TCD) - co-supervisor
Prof. Mariam Kirian - Oak Ridge National Laboratory - USA
https://scholar.google.com/citations?user=QlT-EWQAAAAJ&hl=en
Prof. Albert Cabellos Aparicio - UPC Barcelona Tech
https://scholar.google.es/citations?user=G5f_Mp8AAAAJ&hl=en
Prof. Rafael Pasquini - UFU
https://scholar.google.com.br/citations?user=Qt7oyS0AAAAJ&hl=pt-BR
Prof. André Pacheco - UFES
https://scholar.google.com.br/citations?user=OVhpuAgAAAAJ&hl=pt-BR