Development of a Generative AI-Assisted Network IDS for Intelligent Cloud Cybersecurity Monitoring

In the cloud security paradigm, threat detection and response exhibit structural and functional symmetry, where each detected threat triggers a corresponding automated or manual response. Cloud security is critical due to the increasing reliance on cloud computing to store, process, and transmit sensitive data across various sectors. In order to resist insider threats and external attacks to cloud systems, the present study examines features of DDoS attacks detection and classification based on CICDDoS2019 data set consisting of 88 features and millions of records. Preprocessing was done to eliminate categorical and duplicate features, deal with infinite values, label encoding and min-max normalization. Principal Component Analysis (PCA) was used to select features and therefore dimensionality was maintained by including vital information. An effective model of complex data was acquired by training a Variational Autoencoder (VAE) model (made up of an encoder and decoder) to maximize the evidence lower bound (ELBO). The model had 99.79% accuracy, 98.57% precision, 98.55% recall and 98.56% F1-score, which was better than classical machine learning methods such as RNN, Kalman Backpropagation Neural Networks, and Logistic Regression proving good, dependable, and efficient cloud-based cybersecurity threat detection.