Federated Traffic Classification

Federated approaches to traffic classification: models trained across several networks without sharing their raw traffic, that keep learning new apps and can be explained.

Description

Federated learning trains a shared model across several networks or organizations without moving their raw traffic, which often cannot be shared for privacy or business reasons.

The group applies it to traffic classification:

encrypted traffic classification, with explainable and class-incremental federated models that keep learning new apps;
synthetic traffic generation, training generative AI models in a federated way to produce traffic for classifiers.

Federated intrusion detection is described in Federated Learning NIDS.

People

Antonio PescapèFull Professor · Principal Investigator
Francesco CerasuoloPostdoc
Domenico CiuonzoAssociate Professor
Gabriele MangiacaprePhD Student
Alfredo NascitaAssistant Professor (RTDA)

Publications

Conference G. Mangiacapre, F. Cerasuolo, A. Nascita, D. Ciuonzo, A. Pescapè, “Synthetic Network Traffic Generation via Federated Generative AI for Traffic Classification”, NOMS 2026-2026 IEEE Network Operations and Management Symposium, 2026.
Journal R. Carillo, F. Cerasuolo, G. Bovenzi, D. Ciuonzo, A. Pescapè, “Explainable federated class incremental learning for Encrypted Network Traffic classification”, Computer Networks, 2025.

← All research topics