Enhancing cyber security in WSN using optimized self-attention-based provisional variational auto-encoder generative adversarial network

Abstract

Wireless sensor network (WSN) is a multi-hop and self-organizing wireless network consists of fixed or moving sensors, this is one of the key components of the cyber physical system. It jointly senses, gathers, analyses, and transfer the data of detected objects in the network's service area before sending this data to the network's owner. The attacks like, Black hole, Gray hole, Flooding, scheduling are the usual WSN attacks that could quickly harm the system. A significant level of redundancy, network data's higher correlation, intrusion detection schemes for wireless sensor networks also have the drawbacks of poor identification rate, high computation overhead, and higher false alarm rate. MethodsInitially, the data's are taken from WSN-DS. In pre-processing, it confiscates the data redundancy and missing value restore sunder Color Wiener filtering (CWF). In feature selection, the optimal features are selected using tasmanian devil optimization (TDO) algorithm. Based on the optimum features, the intruders in WSN data are categorized into normal and anomalous data utilizing SAPVAGAN. Hence, honey badger algorithm (HBA) is proposed to optimize the SAPVAGAN, which detects the WSN intrusion accurately. ResultsThe proposed technique is performed in Python utilizing the WSN-DS dataset. Here, the performance measures, like recall, precision, f-measure, specificity, accuracy, RoC, computation time is evaluated. The proposed method provides 23.56%, 12.64%, and 15.63% higher accuracy, 23.14%, 16.78% and 20.04% lower computational time analyzed to the existing models, such as Intrusion Detection System in wireless sensor network using light GBM method (ECS-WSN-SLGBM), Intrusion Detection Scheme in wireless sensor network utilizing recurrent neural network (ECS-WSN-RNN) and Intrusion Detection Scheme for Wireless Sensor Networks utilizing whale optimized gate recurrent unit (ECS-WSN-WOGRU) respectively. ConclusionIt combines advanced techniques such as self-attention, provisional learning, and generative adversarial networks. By leveraging self-attention, the model captures important features and relationships in the WSN data. The provisional allows the model to adapt to changing network dynamics. The component generates realistic sensor data and accurately identifies malicious inputs. Overall, this innovative approach improves security and adaptability in WSNs.