<p>Over the past ten years, Wireless Sensor Networks (WSNs) and the Internet of Things (IoT) have done quite well at enabling real-time applications in a wide range of settings. They are great for large-scale use since they are easy to scale, cost-effective, and can be set up independently. The quick growth of IoT devices and their quick use in WSNs with limited resources, on the other hand, have made these systems more vulnerable to cyber assaults. To make sure that IoT-enabled WSNs are reliable, safe, and profitable, intrusion detection and prevention systems (IDS/NIPS) are very important. This paper presents the Knowledge Distilled Deep Learning Intrusion Prevention (KD-DLIP) framework, a prototype aimed at enhancing the resilience of Wireless Sensor Networks (WSNs) against intrusions in Internet of Things (IoT) settings. The framework uses a teacher-student model, with a teacher model based on a Gated Recurrent Unit (GRU) that finds patterns in network traffic and a lightweight Custom Deep Neural Network (CDNN) student model that learns these patterns via knowledge distillation. The addition of CDNN makes the implementation compact and efficient, which makes the system scalable and good for WSN nodes that don’t have a lot of resources. Experimental assessment shows that KD-DLIP does a better job of stopping intrusions than GRU and CDNN models on their own. KD-DLIP had an accuracy of 0.9919, a macro average F1-score of 0.9918, and a weighted average F1-score of 0.9919. This was better than GRU (0.9877 accuracy) and CDNN (0.9869 accuracy). The model also has better accuracy and recall for both benign (0.0) and malicious (1.0) classifications, which shows that it is strong and dependable. The KD-DLIP framework is a great way to combine the best parts of deep learning and knowledge distillation to provide a scalable, efficient, and accurate way to stop intrusions in IoT-enabled WSNs. KD-DLIP increases the security and reliability of WSNs by obtaining greater accuracy and performance than current methods. This is a big step toward protecting IoT ecosystems from new cyber threats.</p> Graphical Abstract <p></p> <p>Graphical abstract descriptions: This study introduces the Knowledge Distilled Deep Learning Intrusion Prevention (KD-DLIP) framework, an innovative methodology aimed at safeguarding Internet of Things (IoT)-enabled Wireless Sensor Networks (WSNs). Because they are scalable, cost-effective, and used in many sectors, such as healthcare, smart cities, and industrial monitoring, WSNs are now an important aspect of real-time applications. But since there are so many IoT devices and WSN nodes don’t have a lot of processing power or energy, they are very easy targets for advanced assaults. To solve this important problem, the KD-DLIP framework uses a teacher-student learning model. In this model, a strong teacher model made up of Gated Recurrent Unit (GRU) layers learns deep traffic patterns and then passes this information on to a lighter student model through knowledge distillation. This transfer lets the student model keep its high accuracy and dependability while being efficient enough to be used in places with limited resources. The framework also includes Custom Deep Neural Networks (CDNNs) to make sure that the implementation is small without losing any detection capacity. A lot of testing shows that KD-DLIP is better than other intrusion prevention approaches since it is more accurate, scalable, and stable. The suggested framework offers a realistic and robust approach to enhance IoT-based WSN security in forthcoming intelligent applications by facilitating real-time identification and prevention of harmful behaviours.</p>