An intrusion detection system looks through network data to find both legitimate and malicious activity. This study can detect new attacks, which is especially useful in IoT situations. Deep Learning (DL) has demonstrated its superiority in solving challenging real-world issues such as NIDS. This method, however, necessitates more processing resources and takes a lengthy time. During a classification process, feature selection is critical in selecting the best attributes that best describe the goal concept. A novel Network intrusion detection (NEST) technique has been proposed to develop an improved edge-based hybrid feature selection approach, which is a deep learning method for detecting malicious traffic. The Enhanced BPSO technique overcomes the difficulty of BPSO feature selection by combining Binary Particle Swarm Optimization (BPSO) and correlation– based (CFS) traditional statistical feature selection. Three intrusion detection module having three classifiers make up the proposed system. The Signature Detection Module (SDM) examines threats and classifies it as unknown, normal, or intruder based on matching signatures utilizing the Generalized Suffix Tree (GST) algorithm.