Malicious Traffic Classification Using Convolutional Neural Network
Nitin Choudhury, Deepjyoti Deka, Aryan Tewari, S.K. Gaur, Satyajit Sarmah, Vangmayee Sharda, Sitesh Gautam · 2023
While the Internet of Things (IoT) continues to see tremendous expansion around the globe, its security continues to lag far behind. IoT devices were the focus of more than thirty percent of all mobile network spam that were discovered. The use of Neural Networks for the detection of malicious traffic on these often-unattended devices has shown encouraging results. This project is primarily aimed to create an automated system for classification of malicious traffic in an IoT network. In this process, the first step is to prepare/collect a proper dataset for the specific task. The CIC IoT Dataset was used in this instance. The second step includes the preprocessing of the data. In the third step, in order to visualize the files, the files are converted into images using a binary visualizer that is capable of converting any binary file into a colored image. After visualizing the images, the final dataset is prepared. In the fourth step, a Convolutional Neural Network is created and the dataset is fed to the neural network model with 90% training size, 10% validation size of the dataset. During training the model, a 99.9% of training, 99.8% of validation and 81% testing accuracy is obtained. The training loss is reduced to 1.3×10-8and validation loss is reduced to 1.6×10-8.