Taming Volatility: Stable and Private QUIC Classification with Federated Learning

Richard Jozsa 1 , Karel Hynek 2 5 , and Adrian Pekar 1 3 4 · 2025

Read Taming Volatility: Stable and Private QUIC Classification with Federated Learning by Richard Jozsa 1 , Karel Hynek 2 5 , and Adrian Pekar 1 3 4 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-09997.txt.

Description

Richard Jozsa and 2 co-authors study private network traffic classification in this 2025 paper. Federated learning lets many sites train one shared model without pooling raw data, which suits sensitive traffic records. In practice, daily rises and falls in traffic make each site's data come and go, and that uneven timing can destabilize training — a problem earlier work largely overlooked.

Using a real-world QUIC traffic data set split across fourteen clients, the team first shows how unstable standard federated training becomes. They then add a simple client-side data buffer that steadies local training against live traffic swings. The buffered system converged well and scored close to a centralized, non-private model, which the authors offer as a blueprint for stable federated network tools.

Catalog record

CREATORRichard Jozsa 1 , Karel Hynek 2 5 , and Adrian Pekar 1 3 4
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-09997

Browse the full library · About the Tower of Babel