Harnessing feature pruning with optimal deep learning-based distributed denial of service cyberattack detection on IoT environment

Abstract

The Internet of Things (IoT) sensor networks are interconnected networks of physical equipped with sensors, communication features, and actuators. These devices can gather, transmit, and share information with central systems. A Distributed Denial of Service (DDoS) attack is a recent cyberattack that has presented considerable losses in IoT networks. DDoS attacks present an essential threat to operational integrity and network security. While IoT sensor networks accept their increase through several productions, the vulnerabilities to cyber-attacks and malevolent actions improve. Attack recognition using deep learning (DL) utilizes neural networks (NN) to inspect composite designs and abnormalities in data, allowing the recognition of possible security threats. The DL methods like recurrent neural network (RNN) and convolutional neural network (CNN) outshine learning composite feature extraction and data representations, making them appropriate to identify advanced attacks in various areas, like cybersecurity. This manuscript proposes the Harnessing Feature Pruning with Optimal Deep Learning DDoS Cyberattack Detection (HFPODL-DDoSCD) approach. The HFPODL-DDoSCD approach is for efficient and accurate detection of DDoS attacks in IoT environments. The presented HFPODL-DDoSCD approach primarily performs Z-score normalization using a data normalization approach to standardize the input data for enhancing model stability and performance. Furthermore, the Siberian tiger optimization (STO) method is utilized for feature selection, reducing computational complexity while retaining essential information. For the detection of cyberattacks, an integration of self-attention with the bi-directional temporal convolutional network and a bi-directional gated recurrent unit (SA-BiTCN-BiGRU) model is employed. To ensure optimal performance of the SA-BiTCN-BiGRU model, parameter tuning uses the artificial protozoa optimizer (APO) method, fine-tuning hyperparameters to enhance detection capabilities. A comprehensive experimental study is conducted to highlight the significance of the HFPODL-DDoSCD methodology. The comparison study of the HFPODL-DDoSCD methodology portrayed a superior accuracy value of 99.52 % over existing models.

Keywords

• DDoS attack
• IoT
• Cyberattack
• Feature Pruning
• Artificial Protozoa Optimizer
• Deep Learning