Hybrid Machine Learning Model for Feature Selection in DDoS Attack Detection in Cloud Environments Using Convolutional Neural Networks and Genetic Algorithms

As cloud environments are very vulnerable to such threats and are in focus for sophisticated cyberattacks, including the DDoS attack, there is a rising trend in distributed networks as the popularity of cloud computing increases. These attacks usually employ techniques like botnets, spoofing, and multi-vector attacks, and therefore are making becomes increasingly difficult to detect. This paper also presents an adaptive hybrid AI model that uses CNN and GA to select the features of the best option and the detection of DDoS attack. Thus the GA is applied to optimize the feature selection over network traffic and the CNN is then used to learn and extract the spatial and temporal patterns. Subsequently, for testing, our approach undergoes normalization, dimensionality reduction, and feature extraction before the model is tested on CIC-IDS-2017 and CIC-Darknet-2020 datasets. We present a new AI model that performs better than pure AI models like SVM, random forests, and decision trees and surpasses other hybrid methods like HMM and LSTM with a 99.98% detection rate. Results have shown the effectiveness of the proposed model in terms of scalability, reliability and operational ability to work in real time, which can further implied that it is a potential solution for DDoS attack in cloud. In future work, we plan to enhance this framework to include a wider array of attack methods as well as refine the efficiency of the algorithm for use in real time applications with limited computing capabilities.