Design of a Novel Deep Learning Methodology for IOT Botnet based Attack Detection

Abstract

The hackers take advantage of the rapid expansion of the Internet of Things (IoT) to launch attacks on connected devices. There must be a reliable way to identify hostile attacks in order to lessen difficulties about the security of IoT devices. IoT devices are susceptible to botnet attacks, which are common and very hazardous. Static Internet of Things devices are susceptible to security breaches due to a lack of memory and computation results for a platform. Furthermore, there are a number of ways to find new trends in IoT networks in order to provide security. A Recurrent Neural Network (RNN) based on Bidirectional Long Short-Term Memory (BLSTM) is used to build a detection model using as an innovative Deep Learning application. As soon as it detects text, word embedding turns attack packets into integers via tokenization. We compare the BLSTM-RNN detection model to an LSTM-RNN in terms of accuracy and loss for four different routes used by mirai botnet attacks. While the bidirectional approach reduces the processing time and costs every epoch, the study claims that it eventually becomes the better progressive model.