ABSTRACT Wireless sensor networks (WSNs) are crucial for numerous industrial and commercial applications, driven by the rapid growth of Industry 4.0, advancements in wireless technology, and the Internet of things (IoT). However, the throughput and efficiency of WSNs are limited due to the limited lifetime of battery‐operated sensors. Thus, optimal clustering is needed to enhance network performance. This paper presents a novel fuzzy‐based clustering scheme (IFSTA) using an improved shuffle frog leaping algorithm (SFLA) based on Tasmanian devil optimisation (TDO) and an analytical hierarchical process (AHP) algorithm. The TDO is used to enhance convergence, solution diversity and the balance between exploitation and exploration. The IFSTA utilises residual energy, the energy GINI coefficient, inter‐cluster distance (ICD), intra‐cluster distance (ICD), load balancing, coverage and connectivity for optimising the cluster head (CH). The outcomes of the IFSTA are assessed based on network throughput, network lifetime, and residual energy. Further, the deep convolution neural network and long short‐term memory (DCNN–LSTM)‐based framework is utilised for malicious node detection to enhance security. The results show that the IFSTA helps achieve higher network lifetime, throughput, packet delivery ratio and scalability compared with the existing clustering optimisation techniques. The IFSTA provides a 16.38%–51.37% improvement in delay and a 16.37%–167% improvement in network lifetime compared to traditional techniques. The proposed DCNN–LSTM framework achieves an overall accuracy of 98.80%, an F 1‐score of 99.29%, a recall of 99.90% and a precision of 98.80% for malicious node detection on the SensorNetGuard dataset, demonstrating a significant improvement over traditional techniques.
Gunjal et al. (Thu,) studied this question.