Retracted: Online Robustness Model for Intrusion detection model for IP Based Ubiquitous Sensor Network

A group of sensing and technological advancements called wireless sensor networks (WSNs) are used in both defense and commercial research fields. Unattended, distant workspaces with randomized and hierarchically designed sensor networks. Because sensors are placed in an open area and communicate wirelessly, wireless sensor networks are vulnerable to threats. WSNs are susceptible to a variety of security risks, including sinkhole, worm, relaying, grey hole assaults, Sybil, and sink hole threats. Optimal gated recurrent unit (O-GRU) is presented for intrusion detection, which uses packets payload and header information. For both real-time and non-real-time packets are considered for intrusion detection. Furthermore, the proposed method is simulated using network simulation tool. Furthermore, using a deep learning algorithm has drawbacks, including relatively poor accuracy and a reliance on manual feature selection. These issues have been examined, and the current work suggested an improved empirical-based principal components analysis to choose pertinent features. To maintain the majority of the pertinent characteristics, our presented feature selection methodology combines the benefits of both empirical mode decomposition as well as principal component analysis. Machine learning model was used to classify the attack node based on the chosen characteristics (Long Short-Term Memory). In comparison to state-of-the-art techniques, the proposed model verified the CICIDS 2017, NSL-KDD, UNSW NB 2015, and KDD99 datasets. The efficiency of the presented system was demonstrated by comparison with both the widely used approaches.