Multiple Attack Classification in IoT Networks

IoT technologies are increasingly emerging to include almost all aspects of life in healthcare, agriculture, industry, transportation, and many more fields. However, IoT technology is known to have weak security, which makes it an easy target for malicious users. Security research in IoT can be split into two phases protection and detection. Attack detection is a vital stage in network security. In this paper, we present a machine-learning approach for attack detection in IoT networks together with attack classification. We include four different machine learning models, CNN, LSTM, RandomForest and a hybrid model, CNN-LSTM. The model’s performance was tested and compared in three various train-to-test portions. The results state that RandomForest performed the best compared to other models, with 99.976% accuracy. The hybrid CNN-LSTM model improved the performance of CNN, increasing accuracy by 6.85%, and reducing the error rate in regular traffic by 40.26%.