Efficient Detection of DDoS Attacks Using a Hybrid Deep Learning Model with Improved Feature Selection

Daniyal Alghazzawi,Muhammad Zubair Asghar,Hayat Ullah,Omaimah Bamasag

doi:10.3390/app112411634

Daniyal Alghazzawi, Muhammad Zubair Asghar + Show 2 more

Open Access

https://doi.org/10.3390/app112411634

Copy DOI

Journal: Applied Sciences	Publication Date: Dec 8, 2021
Citations: 40	License type: CC BY 4.0

Affiliation: King Abdulaziz University, Gomal University

Abstract

DDoS (Distributed Denial of Service) attacks have now become a serious risk to the integrity and confidentiality of computer networks and systems, which are essential assets in today’s world. Detecting DDoS attacks is a difficult task that must be accomplished before any mitigation strategies can be used. The identification of DDoS attacks has already been successfully implemented using machine learning/deep learning (ML/DL). However, due to an inherent limitation of ML/DL frameworks—so-called optimal feature selection—complete accomplishment is likewise out of reach. This is a case in which a machine learning/deep learning-based system does not produce promising results for identifying DDoS attacks. At the moment, existing research on forecasting DDoS attacks has yielded a variety of unexpected predictions utilising machine learning (ML) classifiers and conventional approaches for feature encoding. These previous efforts also made use of deep neural networks to extract features without having to maintain the track of the sequence information. The current work suggests predicting DDoS attacks using a hybrid deep learning (DL) model, namely a CNN with BiLSTM (bidirectional long/short-term memory), in order to effectively anticipate DDoS attacks using benchmark data. By ranking and choosing features that scored the highest in the provided data set, only the most pertinent features were picked. Experiment findings demonstrate that the proposed CNN-BI-LSTM attained an accuracy of up to 94.52 percent using the data set CIC-DDoS2019 during training, testing, and validation.

Highlights

DoS (Denial of Service) attacks diminish a particular system’s network bandwidth and computational resources by overloading it with malicious traffic, blocking it from providing normal services to authorized users
Our goal is to develop an automated approach that learns from provided training data to anticipate Distributed Denial of Service (DDoS) attacks using a hybrid deep neural network model with optimized feature selection
For the sake of being practically useful, this study addresses the problem of DDoS attack detection by adding optimum feature selection into the proposed hybrid deep learning (DL)-based architecture

Summary

Introduction

DoS (Denial of Service) attacks diminish a particular system’s network bandwidth and computational resources by overloading it with malicious traffic, blocking it from providing normal services to authorized users. DDoS (Distributed Denial of Service) [1] takes things a step further on a wider scale. Distributed Denial of Service (DDoS) attacks are DoS attacks that are executed in a distributed way to increase the resource usage for one or more targets [2]. DDoS attacks seize control of the majority number of compromised systems, known as a botnet, and execute coordinated attacks on the target machine. DDoS attacks are developing and increasing in magnitude, frequency, and complexity in tandem with the introduction and growth of innovative Web-based technologies. Companies confront possible network risks that might have serious consequences for their activities, such as outages, data theft, or even blackmail threats from cybercriminals [3]

Objectives

Methods

Findings

Conclusion