The statement is true in the context of Huawei’s AI-assisted application-identification architecture. Conventional identification depends mainly on predefined signatures, fixed destination addresses, port numbers, protocol fields, DNS correlation, or deep packet inspection. Huawei’s standard SD-WAN process uses Service Awareness and First-Packet Identification signature databases to identify and group application traffic. These mechanisms are effective for known applications but become less reliable when traffic is encrypted, applications change versions, or previously unseen applications appear.
A CNN-based classifier learns multidimensional traffic characteristics such as packet-length sequences, timing, direction, and flow behavior. It can classify traffic matching learned application patterns and, when implemented with open-set detection, recognize flows outside known classes as unknown or zero-day applications. Commercial-grade deep-learning traffic classification research demonstrates identification of known encrypted applications together with the handling of unknown zero-day applications.
This does not mean the system automatically assigns an exact commercial name to every unseen application. It detects that the traffic does not match established classes so that it can be investigated, labelled, and incorporated into later model updates. Therefore, the statement is True.
==================
