Classification of Intrusions in RPL-Based IoT Networks: A Comparison

Internet of things is a group of objects or devices that establish communication through Internet. Since, the communication is through Internet, more processing power is required in the devices. The IoT devices are heterogeneous and energy constrained. Owing to constrained processing power, the traditional security procedures are not applicable to be executed in the devices. This generates different attacks like user to root, DoS, remote to local and probe attacks. In the existing systems, the classification is carried out to classify different intrusions by using various machine learning algorithms based on KDD dataset. The results were analyzed with various parameters. ML algorithms led to false classifications with medium accuracy. So, it was unable to classify various attacks perfectly. In this paper, the two variants of recurrent neural network such as long short-term memory and bidirectional LSTM are used for the classification of different IoT attacks. The proposed algorithms reduced the false classification with good accuracy. The parameters are computed for the proposed algorithm and are compared with existing machine learning algorithm.