Deep Learning based Intrusion Detection System for WSN

Wireless sensor networks (WSNs) have so many uses, both in the military and in the civilian world, they are becoming one of the most popular research topics in computer science. Many sensor nodes make up a WSN, and they all gather and forward data to the central point. However, there are a number of security concerns with WSN due to deployment strategies, communication routes, and resource-constrained nodes. Determining unauthorized access is therefore crucial to enhancing security. The Intrusion Detection System (IDS) is necessary to guarantee the dependability and security of WSN services. This IDS must be able to identify the greatest number of security threats and be in harmony with the features of WSNs. All communication networks require the services provided by the network intrusion detection system (NIDS). IDS frequently include machine learning (ML) approaches; yet, ML techniques’ effectiveness is subpar when dealing with imbalanced attacks. This research proposed an IDS implied on deep neural networks (DNN) to enhance performance. The most effective features from the dataset are chosen through the cross-correlation method. Next, a DNN structure that is intended to detect intrusions is developed using the parameters that have been chosen. The NSL-KDD intrusion dataset is used for testing-training for the proposed model. According to the experiment’s outcomes such as Accuracy 96.23%, Precision 95.75%, and Recall 92.82% the proposed DNN performs more efficiently in detecting attacks over traditional ML models like RF(random forests), DT(Decision trees), and SVM(Support vector machine) and DL model like Autoencoder.