Enhanced cybersecurity solution via distributed denial of service attack recognition and attribute subset selection

In the cybersecurity domain, a drastic increase in threats affects the security of network and users. Particularly, Distributed Denial of Service (DDoS) attack poses a challenging issue because of its ability to inundate and debilitate targeted systems via distributed networks. Recently, deep learning (DL) based intrusion detection systems (IDS) are employed for detecting network intrusions and identifying anomalies in the network traffic. This study proposes a Cybersecurity Solution for Attack Detection Using Dimensionality Reduction and Multiple Classifier Systems (CSAD-DRMCS) methodology to enhance cybersecurity against DDoS attacks in advanced network infrastructures. The presented model applies dimensionality reduction by the use of Fisher Score, mutual information (MI), and information gain (IG) methods to choose optimal set of features. Moreover, ensemble classification model is developed by the integration of temporal convolutional autoencoder (TCAE), Elman neural network (ENN), and deep belief network (DBN) to increase intrusion detection capability and classification robustness. The simulation analysis of the proposed model is tested using the Edge-IIoT and CICIDS2017 datasets highlights the betterment of the proposed CSAD-DRMCS framework, achieving maximum accuracy values of 99.19% and 99.21%, respectively. The accomplished results assured that the proposed model effectively improves DDoS attack detection performance, enhances cybersecurity reliability, and provides a scalable and intelligent solution for securing modern network infrastructures.The source code supporting the findings of this study is available in the GitHub repository at: https://github.com/zgenz1537-code/UCECEarlyDetection.git .