Distributed Denial of Service Attack Detection in IoT Utilizing Attention Mechanism Based Long Short-Term Memory

Internet of Things (IoT) is a highly impactful approach which has become ubiquitous in our daily lives, particularly when it comes to safeguarding user data and personal information. Protecting the IoT infrastructure with a traditional Distributed Denial of Service (DDoS) is a highly challenging task due to the vast variety and number of IoT devices. This research proposes the Attention-based Long Short-Term Memory (A-LSTM) for DDoS attack detection in IoT system. The proposed A-LSTM method utilized the two IoT datasets named Bot-IoT and UNSWNB15 for estimate the performance. In this research, a pre-processing step is performed for handling the missing values and normalization in the collected dataset. Then, pre-processed data is selected by using Particle Swarm Optimization (PSO) approach. The A-LSTM is utilized to classify DDoS attack into malicious or normal. The proposed A-LSTM approach accomplishes superior results like accuracy of 99.72 and 97.91% in both Bot-IoT and UNSWNB15 dataset respectively when compared to the previous approaches named Deep Neural Network (DNN), Feedforward Neural Network (FNN) and LSTM.