Automatic detection of intrusion attacks in iot networks usingBI-LSTM-CNN neural network

In today's fast-paced technological landscape, the rise in cyber attacks and unusual incidents highlights the urgent need to quickly spot these irregularities, which can lead to various problems. The problem attacks and anomalies in Internet of Things (IoT) networks are more paramount; the significance of it is manifested by the criticality of the applications served by the network through the interest. Power networks, large manufacturing industries, etc. Will face a severe threat to any irregulations in IoT network. These attacks can take many forms, including viruses, hacks, disruptions, etc. That's where anomaly detection comes in – it's a way to carefully identify these unusual occurrences in data streams that stand out from the norm. To tackle this challenge, we suggest a new approach that combines two techniques: bidirectional long short-term memory (Bi-LSTM) and convolutional neural networks (CNN). This pairing has the potential to make identifying anomalies faster and more accurate. Our method is built on the well-established UNSW-NB15 dataset, which we've tested extensively in a controlled cyber environment using specialized tools. To make our model even better, we've added a technique using rectified linear weights. These weights are specially designed for the Bi-LSTM with CNN setup and provide improvements over the standard weights. We have compared our method with existing ones, and our results are impressive. Our model achieves a detection accuracy of 94.14%, surpassing by 10.18%, by 13.89%, and by 14.40%. This means our approach is highly effective in quickly and accurately identifying anomalies, contributing to enhanced cybersecurity.