Energy-Efficient Intrusion Detection in WSN-IoT Using Modified Dingo Optimization and Fuzzy Adaptive DeepNet

Wireless Sensor Networks (WSN) and Internet of Things (IoT) need effective intrusion detection, which weighs between the high level of detection accuracy, low energy and low false detection rate. We introduce a power-efficient framework that uses Modified Dingo Optimization (M-DO) on fine-grained feature selection and Fuzzy-Improved Fuzzy Adaptive DeepNet (IFA-DN) on end-to-end attack classification. The path to designing the m-DO and IFA-DN is to exclude unimportant attributes to decrease the computational cost, as well as false alarms and complex intrusion patterns, by using the hybrid logic fuzzy and self-attention enhanced DeepNet architecture. Our strategy demonstrates a high degree of reliability and low overfitting through a 96.92 detection accuracy and an AUC of 0.9665 on the UNSW-NB15 and 99.77 as detection accuracy and 0.9971 as AUC score on the NSL-KDD benchmark, which operate in resource-limited settings.