Validation of Deep Learning-based Hybridization Model for DDoS Attack Detection with Performance Metrics Comparison

Abstract

Separating fraudulent from valid traffic is the main difficulty in a Dispersed Denial-of-Service (DDoS) attack. DDoS assaults are deliberate attempts to obstruct any computer, network, or support system from operating normally by flooding the target or nearby resources with an enormous volume of Internet traffic. This type of attack can be a single-source attack or a complicated multi-source attack, among other variations. In this study, a novel deep learning classification method was proposed by hybridizing two common deep learning algorithms; DDoS attack detection using intelligent deep neural unified sequential memory networks (IDNUSMN). The model was tested on the NSL-KDD dataset. Z-score normalization was used as a preprocessing step is used to convert data into standard normal distribution. The proposed method is implemented using Python software. The IDNUSMN is compared to the other traditional algorithms. According to the results, the IDNUSMN outperformed the others in terms of F1-score (98.35%), precision (98.18%), recall (98.5%), and accuracy (98.25%). The study's validation results demonstrate how effective the hybridization model based on deep learning is in detecting DDoS attacks.