Botnet Attack Detection Using Deep Learning in the Internet of Things Networks
Aum Shah, B. Kanisha, Sheetal Jatav · 2024
Botnet assaults on Internet of Things (IoT) devices are extremely harmful, yet deep learning (DL) detection requires a lot of memory and traffic on the network, and is not effective on devices with minimal memory. Dimensionality reduction techniques can be used to reduce the number of attributes in IoT network traffic. Thousands of botnet assault events classified as DDoS, DoS, reconnaissance, and information stealing are included in the publicly accessible Bot-IoT dataset. The IoT network can utilize the dataset to detect botnet traffic. Two-dimension reduction methods that might be helpful in reducing the dataset’s total feature dimensions are PCA and autoencoding. An intermediary hidden layer is used by the autoencoder, an unsupervised deep learning method, to produce a latent-space representation of the data. Deep learning techniques, such as Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN), can detect botnet attacks using the decreased feature set. The approach’s success can be assessed using performance metrics such as accuracy, precision, recall, and a confusion matrix. In conclusion, the suggested approach is to lower the dimensionality of the features of the Bot-IoT dataset using dimensionality reduction techniques like PCA and autoencoder, enabling DL models like CNN and LSTM to detect botnet attacks. There are several performance measures available for assessing the technique.