Network Traffic Attack Detection using Dimensionality Reduction based Dual Strategy-Horned Lizard Optimization algorithm in Internet of Things

Abstract

The network traffic detection for Internet of Things (IoT) includes monitoring and analyzing data traffic patterns to recognize malicious activities like Denial-of-Service (DoS) which helps to ensure the security of IoT networks. However, enhancing volume of network traffic in IoT system makes it difficult to detect cyberattacks accurately because of huge number of features which most are redundant led to inefficiencies in detection performance. To overcome this issue, Dual Strategy- Horned Lizard Optimization Algorithm (DS-HLOA) is proposed for improving attack detection through applying optimization-based feature selection techniques which removes redundant features and enhance accuracy. The strengthened convergence and mutation are the di strategy used to identify the optimal solution and avoid the local optima. The Long Short-Term Memory (LSTM) performed network, detecting traffic attacks through capturing long-term dependencies in sequential data for identifying complex pattern in traffic data. The DS-HLOA achieves accuracy of 98.21%, 96.99% and 99.98 % for BOT-IoT, CICIDS 2018 and 2017 when compared to Rat Swarm Hunter Prey Optimization- Deep Maxout Network (RSHPO-DMN) and hybrid machine learning approach.