Harmonizing Deep Learning and Ensemble Techniquesfor Network Traffic Anomaly Detection
Ayush Kumar Kashyap, Madhumita Mahapatra, Piyush Yadav, Seema Verma · 2024
Keeping not only the security of the network infrastructure, but the innovativeness of the cyber threats in mind, is vital for the safety of the digital landscape today. This study aims to inspect which strategies are more applicable by combining deep learning with ensemble methods for elaboration of network traffic anomaly detection, especially dwelling on the most productive methodologies. Through employing the Train-Test-Evaluation scheme on the CICIDS 2017 dataset, we deployed advanced techniques including Convolutional Neural Networks (CNN), Deep Neural Networks (DNN), Recurrent Neural Networks with LSTM as well as ensemble methods like AdaBoost, Gradient Boosting, Random Forest, Isolation Forest, and Voting. Evaluation metrics which measure performance of algorithms such as Accuracy, Precision, Recall, and Fl Score were used to compare the performance of these techniques. Analysis of our voting method gave us confidence in the described term. It generated the highest in detection accuracy, precision, and recall among others considered models. This work enhances the fields of network security with the presentation of the right detection techniques of the anomalies, the importance of the ensemble methods and the voting classifier in fighting cybercrimes on the network.