CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Detección de ataques mediante aprendizaje federado (FL) en redes IoT

Estudiants que han llegit aquest projecte:

Director/a: LEÓN ABARCA, OLGA

Departament: ENTEL

Títol: Detección de ataques mediante aprendizaje federado (FL) en redes IoT

Data inici oferta: 18-07-2023     Data finalització oferta: 18-03-2024


Estudis d'assignació del projecte:
    GR ENG TELEMÀTICA
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
Internet of Things (IoT);Dataset;Seguridad;DoS;DDoS;Mirai;Ataques Web;Machine Learning;Federated Learning
 
Descripció del contingut i pla d'activitats:
Federated Learning allows training machine learning models with
decentralized data while preserving its privacy by design. Thus,
it appears as an ideal solution to detect attacks to IoT devices
without the need of revealing sensitive information.
 
Overview (resum en anglès):
The continuous increase in the popularity of IoT devices can bring significant advantages, whether in automation, efficiency, data collection, etc. However, as a consequence, both the lack of security in these devises and the sensitive information they may store about users make them one of the primary targets for attackers, either to obtain information or to use them maliciously. To detect these attacks, the most popular tool today, Machine Learning, is initially considered. However, since its operation requires data from each device, it can increase network traffic to the point of saturation. Additionally, sending sensitive data over the network can lead to unwanted information leaks. Then, as an alternative, the Federated Learning technique emerged, allowing each device to apply a Machine Learning algorithm locally. This way, devices only send information about the algorithm is learning process instead of raw data. The main objectives of this project focus on preparing datasets extracted from the CICIOT2023 dataset, which includes samples of natural network traffic in an IoT network and a set of samples from 33 different attacks carried out on various IoT devices. These datasets will be used for both Machine Learning and Federated Learning. The project will also focus on analysing the results provided by Machine Learning and Federated Learning algorithms in detecting threads in IoT networks, determining which algorithm achieves better detection, and comparing the accuracy of Machine Learning and Federated Learning.

© CBLTIC Campus del Baix Llobregat - UPC