No abstract

A wireless sensor network consists of a large number of sensors dispersed across a large area. These are used in broad areas including queue management, military applications, ecological applications, and others. This method, which combines deep learning and optimisation strategies with a focus on attack identification, is still under testing. The nodes will first be distributed randomly, centred on the network's dimension, under a system paradigm. Comparison sets are produced by use of an energy-related timer. Later, the geographical comparison, the quality of the link between the cluster head (CH) and cluster member (CM) nodes, and the node's remaining network energy will all be taken into account when analysing the transmission probability. The CH will determine how to manage the trust. The node will be chosen as CH after it meets the criteria for trust coverage. This will be chosen as CM if the situation is still unsatisfactory. The Dempster-Shaft theory and multi-dimensional trust criteria will be used to determine the cluster pathways' (CP) optimal range for effective data transfer, with residual energy and distance being the key constraints. Cascaded Hermite Laguerre Neural Network will classify and identify the attack if the best and most reliable path is still chosen (CHLNNet). This proposed approach will be compared against three sophisticated methodologies with regard to several parameters. As a result, the suggested CHLNNet technique achieves 91.4% of malicious detection rate, 28.2% average latency, 94.8% throughput, 23% end-to-end delay, and 31.4% routing overhead.