The Internet of Things (IoT) is a largely emerging area having applications in almost all sectors but a threat to the security of the IoT network is the main hurdle in the growth of IoT networks. For attack detection, standard IoT datasets are used. These datasets are highly imbalanced with major benign traffic and very little attack traffic. To deal with the imbalanced dataset in this paper, different resampling techniques such as Undersampling, Oversampling, and hybrid sampling are applied to the CIE-CICIDS2018 dataset. After resampling, the artificial neural network is applied for attack detection on this resampled dataset. As the dataset is imbalanced, for evaluation of the performance along with accuracy, precision, recall, and the F1 score are parameters used. Random Undersampling is giving the best result among all resampling techniques but a lot of data loss occurred in Random Undersampling. Edited Nearest neighbors is giving better results than all other techniques except Random Undersampling without losing the majority of data samples.

class imbalance,CICIDS2018, IDS, resampling,