Artificial Intelligence (AI)-Driven Network Traffic Anomaly Detection for IT Infrastructure Security

Anomalies are common in the monitoring of network systems. It is critical to recognize such anomalies in network behavior when it materialize as network hazards to be mitigated, service outages to be avoided, and security issues to be addressed. To determine some suspicious network traffic on CICIDS-2017 dataset, this study develops a CNN-BiLSTM model of hybrid convolutional neural networks (NN). The proposed architecture will contain systematic preprocessing procedures to guarantee quality input information, such as data cleaning, one-hot encoding, feature selection, minmax normalization, and data balancing. The BiLSTM is used to extract the temporal relations and the CNN portion is used to reconstruct the qualities of the network data in space, enabling the realistic modeling of sequential behaviors that are interrelated to cyberattacks. Experimental findings indicate that the suggested model reached 96.58% accuracy (ACC), 97% precision (PRE), 95.64% recall (REC), and F1-score (F1) of 96.09% which are better than the traditional models including Random Forest, Logistic Regression, LGBM, Autoencoder, CNN-LR, XGB and SVM. The comparison shows that the model is better at detecting various types of network anomalies, which make it an incredibly powerful and scalable next-generation intrusion detection system.