The technological evolution of IoT devices allows for a complexity in networking the data that is produced and created daily. The complexity of net working in terms of data flow traffic via the worldwide web, the Internet of Everything, it needs the powering of CPU and GPU to process rapidly the flow of requests, which they are sent by IoT users. It requires also an intelligent system to optimize Traffic data networking. This system allows us to guarantee the high quality of learning parameters, which play a very important role in the training phase of a learning model. The excellent network traffic engineering allows to ensure that all parameters transmitted or exchanged via the network are reliable, and they are not subject to any type of attack that would allow them to fail to predict satisfactory results expected by IoT devices or users via APIs (Application Programming Interfaces). Network traffic engineering relies on interconnection equipment such as routers, switches, and gateways, which must be intelligently configured. This intelligence is provided by the implementation of new technologies such as SDN, which make the configuration of interconnection equipment more reliable and automated through the use of programs deployed on one of the essential components of this technology: the SDN (Software Defined Network) controller. For this network, we have analyzed the best architectures that deploy a federated learning model in a reliable and secure manner. These architectures attempt to address all kinds of issues related to the security of learning parameters produced by IoT devices during the model training phase, as well as data privacy and confidentiality. The model training phase is carried out at the level of very powerful software and hardware network infrastructures to ensure very high-quality training of a learning model.