Effective Internet of Things botnet classification by data upsampling using generative adversarial network and scale fused bidirectional long short term memory attention model

Abstract

Internet of Things (IoT) botnet attacks are considered an important risk to information security. This work mainly focusing on botnet attack detection targeting various IoT devices. In this work, feature generation and classification are the two major processes considered for attack detection. Generative adversarial network (GAN) is applied for the feature generation process. GAN has generator and discriminator. Here effective generator network is introduced by applying added convolution layers with batch normalization and rectified linear unit activation function. In this proposed system, a novel network called the data perception network is proposed with scale fused architecture. The data perception network is developed to determine generator's efficiency in generating fake data similar to original data. This perception network is also considered for estimating loss function by analyzing in different scales. Hence, the major strength of this network is that highly reliable data are provided using the synthesized data. An efficient network architecture called scale fused bidirectional long short term memory attention model (SFBAM) is applied for the classification process. The proposed model is evaluated using the IoT‐23 dataset, which can differentiate between benign and malicious data in IoT attacks. Compared to existing models, this proposed model provides effective results by improving accuracy and reducing loss.