Past two decades, Wi-Fi innovation has developed so generally that most contemporary gadgets are compact and use Wi-Fi to access the web. Because no physical boundary separates a wireless network from a wired network, Wi-Fi network security is seriously questioned, and current security measures are helpless to an extensive variety of attacks. The objective of this research was to evaluate a novel approach called federated learning as a potential solution to address privacy concerns and high costs associated with data collection in the field of recognizing network attacks. The study introduces FEDDBN-IDS, a ground-breaking intrusion detection system (IDS) that utilizes deep belief networks (DBNs) within a federated deep learning (FDL) framework to detect and identify cyber threats specifically in Wi-Fi networks. A DBN with stacked restricted Boltzmann machines (RBM) was trained on each device to learn the low-dimensional features from unlabelled private and local data. A central server later combines these models into a global model (FL) using federated learning. Subsequently, the central server adds fully connected SoftMax layers to the overall model to create a supervised neural network and train the model using freely available labeled AWID datasets. The results obtained from our experiments on the AWID intrusion detection dataset indicate that our federated strategy achieves a high level of classification accuracy, ranging from 88% to 98%.