AHSWN Home · Issue Contents · Forthcoming Papers

Enhancing Reliable Communication in IoT Assisted Wireless Sensor Networks via Artificial Intelligence Enabled Automated Threat Detection and Classification
R. Gayathri and N. Kumaratharan

The integration of Internet of Things (IoT) devices and wireless sensor network (WSN) technology establishes a network designed to share and process data. Given the volume of private information handled by this data-centric infrastructure, security issues pose a serious threat and require immediate attention. Given the inherent nature and constraints of WSNs, data are highly susceptible to modification, corruption, or theft. Intrusion detection is a proactive security mechanism that monitors the network’s operational status and detects security breaches, enabling the network to mitigate threats and respond quickly. Artificial intelligence (AI), specifically deep Learning (DL), has attained reliability through an effective application in the recognition of network threats, including IoT networks. Due to the resource-constrained nature of WSN nodes, traditional network intrusion detection systems are not viable without modification. Currently, numerous researchers in WSN intrusion detection apply DL techniques to analyse traffic data. DL techniques can minimize the computation burden and raise the capability to learn the traffic data features, which can enhance the accuracy of the detection framework. This research introduces an Advanced Artificial Intelligence-Enabled Threat Detection and Classification for Reliable Communication (AAI-TDCRC) framework for IoT Assisted Wireless Sensor Networks. The main objective of this research is to create a robust security mechanism capable of accurately detecting and classifying threats in IoT assisted wireless sensor network environments, thereby enhancing data integrity and secure communication across the network. In the initial stage, the data preprocessing is carried out through steps like data cleaning, normalization, scaling, and data splitting to ensure data quality, consistency, and format, leading to improved model training. Following that, a three-stage feature selection pipeline is performed, integrating minimum redundancy maximum relevance, ReliefF, and LASSO, which identifies the most relevant features contributing to the detection process. Finally, the multiple graph neural network architectures with a bidirectional long short term memory network are designed for robust threat detection and classification. To ensure the improved performance of the AAI-TDCRC model, an extensive simulation analysis was conducted, and the results indicated the improvement of the AAI-TDCRC model over the existing models.

Keywords: Wireless Sensor Networks; Internet of Things; Threats; Artificial Intelligence; Feature Selection; Hybrid Classification