Feature Selection Using Firefly Algorithm for Classification of Attacks in Routing Protocol for Low Power Lossy Networks (LLNs) Based Internet of Things (IoT) Networks

IoT(Internet of Things) is a collection of diverse resource-constrained things or nodes. All the IoT devices used to collect and share data are connected through Internet. Since all the devices are connected through Internet, it makes the devices to process more data. Standard security methods could not be applied in the devices due to a lack of sufficient resources like processing power. So, RPL is used as a routing protocol in most of the IoT networks. Due to constraint in resources, RPL is vulnerable to various attacks. In existing systems, various Machine Learning (ML) algorithms are employed to classify different IoT attacks. Since ML algorithms are linear, Deep learning (DL) algorithm is used in this work for classification of attacks. Furthermore, Feature selection approach is employed to improve the performance of the classification. Firefly algorithm is used for selecting the features in this work to remove the irrelevant features. The features that increase the accuracy of the model are only considered. The selected features are only given as input to LSTM model. From the result, it is inferred that the accuracy of Long Short Term Memory (LSTM) with feature selection algorithm is 98.53% and without feature selection algorithm is 93.25%.