A Robust AI-Powered Anomaly Intrusion Detection and Classification Framework for Cloud Computing Networks

The widespread adoption of cloud computing services across different industries creates security risks which stem from probable cyberattacks. A comprehensive method for identifying cloud computing network intrusions utilizes the Inception V3 deep learning model and the UNSW-NB15 dataset. The dataset contains over 2.5 million network streams that include nine attack types and benign traffic while going through detailed processing that removes null values, eliminates duplicates, applies one-hot encoding, scales features and selects features to boost model performance. To determine that the classes are balanced or imbalanced and to infer relationships between the features, methods such as pie charts and heatmaps were applied. The Inception V3 model, trained on the above-said dataset, proved high efficiency with 0.984 accuracy, 0.99 precision, 0.99 recall and 0.985 of F1score, which out-competed traditional models like CNN + LSTM, IIDNet, ANN, DT, and CNN. Other graphics, such as the training and validation accuracy, the loss diagrams, and the confusion matrix, also explain how viable the model is for generalization. These results prove that the obtained Inception V3 architecture is very efficient and can be used effectively for identifying multifaceted attack patterns to boost cybersecurity in cloud networks.