IoT Intrusion Detection Using Two-Tier-Convolutional Deep-Learning Model

Security has elevated to a top priority with the expansion of devices connected in the IoT. This research proposes a two-tier convolutional deep-learning model-based IDS to handle this problem. The suggested IDS calls for pre-processing step-by-step min-max normalisation, feature extraction, and use of statistical features like skewness, mean, standard deviation, entropy, and kurtosis, as well as flow-based variables like flow length, flow size, flow rate, and flow direction. In order to train the two-tier convolutional deep-learning model, the extracted features are utilised. An innovative SA-OOA is put out to improve the model's activation function, accuracy and the rate of false positives. The suggested approach adjusts the characteristics of the input layer during training to intensify the performance of the model. The CICIDS2018 database and the model's implementation in MATLAB are used to assess the effectiveness of the suggested IDS. The experimental findings show that, in respect of accuracy rate and false positive rate, the proposed IDS accomplishes more than current state-of-the-art IDS. IoT network intrusion detection is crucial for ensuring the safety of devices connected and the data they communicate, and the suggested IDS offers a practical and effective solution.