1D-CNN-based Real-Time Network Intrusion Detection with Privacy-Preserving for IoT

Abstract

The Internet of Things, or IoT, is a concept that links every device to the Internet and enables them to collaborate to achieve shared goals, including smart automation in the home. The amount of data produced in the rapidly growing Internet of Things (IoT) space has never been higher.  Numerous decision-making processes are accelerated by processing this enormous data to get priceless insights. Intrusion detection systems (IDS) are crucial for safeguarding data and computer resources against external threats in computer networks.  Modern IDSs struggle to become more effective and flexible in the face of unexpected and unexpected threats.  The suggested approach presents a real-time, privacy-preserving; 1D-CNN-based network intrusion detection system for the Internet of Things. This paper introduces an Intrusion Detection System (IDS) that utilizes a dataset to enhance the Internet of Things (IoT). First, the input data is preprocessed using data processing. Due to its ability to decrease convergence, the choice of features is included in the suggested model used to improve linearity-based principal components analysis. Intrusion detection is recognized using algorithms with extreme gradient boosting (XGBoost) hyperparameters.  1D CNN is the best and least costly model for real-time IoT security monitoring, and it excels in the recommended approach of feature extraction. With a binary intrusion detection accuracy of 99.77%, 1D CNN again outperformed LSTM, RNN, and MLP, which obtained 98.25%, 94.52%, and 92.2%, respectively.   The results, based on feature extraction, demonstrate that 1D CNN is a highly efficient real-time IoT security monitoring technique. Among all models that are used for analysis and development, it offers the best efficiency and dependability. Lastly, the performance of the proposed model was superior to that of the existing models.