Artificial Intelligence-Powered Network Intrusion Detection System (IDS) with Hybrid Deep Learning Approach in Cloud Environments

Nowadays, more and more people are becoming aware of the security issues of computers. Traditional IDS (Intrusion Detection Systems) have the drawbacks of poor realtime performance and low accuracy. In response to this issue, this paper adopts an ML-based network intrusion detection algorithm. The system utilizes the CSE-CIC-IDS2018 dataset, which undergoes rigorous preprocessing, including duplicate removal, missing value handling, and Min-Max normalization to ensure high-quality input. Dimensionality reduction is performed using PCA techniques to enhance learning efficiency. A hybrid deep learning based LSTM-Petri Net (LSTM-PNet) architecture is employed with Adam optimizer and binary cross-entropy loss function as utilized hyperparameters. Training accuracy was 99.70%, and validation accuracy was 99.68%, both with minimum loss values, demonstrating exceptional performance of the proposed model. Classification metrics indicate near-perfect precision, recall, and F1 scores across both benign and attack classes, supported by clear convergence in accuracy/loss curves and an interpretable confusion matrix. When benchmarked against traditional and deep learning models like Decision Tree, DNN, CNN, AdaBoost, and MLP-PSO, the proposed model consistently outperforms all others, underscoring its robustness and adaptability.