CloudETC: a Privacy-Preserving Encrypted Traffic Classification Platform for QoS in Wi-Fi

Traffic classification (TC) is essential to many Quality of Service (QoS) provisioning frameworks. Most modern network traffic is encrypted with Transport Layer Security (TLS) protocol, and usually, TC is carried out by analysis of the TLS parameter server name indication (SNI). This parameter contains information about the host a client tries to connect to. However, the TLS workgroup has protected SNI in the TLS Encrypted Client Hello (ECH) amendment, and only complex and continually updated TC algorithms can achieve sufficient classification accuracy in scenarios with ECH. Yet, many home and office Wi-Fi access points (APs) are short on computational resources, and the owners rarely upgrade their software. This demo addresses this challenge with the Traffic-Classification-as-a-Service concept. The idea is to send several first packets of the flows to a cloud server that classifies them and suggests an appropriate QoS policy to the AP. Notably, we preserve privacy by sending only the first packets that do not contain users' payload. The demo confirms that this concept may double the bitrate for YouTube traffic in a highly loaded network.