Banca de QUALIFICAÇÃO: DANIEL WALMIR DOS SANTOS ALVES

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : DANIEL WALMIR DOS SANTOS ALVES
DATE: 25/11/2025
TIME: 09:00
LOCAL: remoto
TITLE:

Permutation Feature Importance in Federated Learning-based Intrusion Detection Systems for IoT Networks

KEY WORDS:

Federated Learning; Permutation Feature Importance; Internet of Things; Intrusion Detection; Network Security

PAGES: 62
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUMMARY:

The continuous growth of the Internet of Things expands the cyber-attack surface while naturally constraining energy, processing power, and bandwidth at the network edge, which challenges the direct application of traditional Intrusion Detection Systems. This research proposes integrating the Permutation Feature Importance (PFI) metric into the Federated Learning (FL) cycle as a dynamic feature-selection strategy to reduce communication overhead among participants and preserve data privacy. The proposed architecture is organized into three layers: end devices, edge gateways, and an aggregation server. Enabling the coordination of local models within a distributed and secure environment. The methodology encompasses a systematic literature review, the conceptual design of the architecture, the experimental implementation of a reproducible prototype, and the planning of tests to evaluate network usage, convergence time, and detection performance. The study is expected to deliver an open artifact that combines adaptive compression, interpretability, and distributed defense, contributing to the development of lightweight and transparent IDS solutions applicable to diverse Internet of Things scenarios.

COMMITTEE MEMBERS:
Presidente - 3139050 - ROGER KREUTZ IMMICH
Interno - 1874895 - RAMON DOS REIS FONTES
Externo à Instituição - DIEGO LUIZ KREUTZ
Externo à Instituição - RODRIGO DA ROSA RIGHI