Leveraging Deep Learning and Feature Extraction for Robust Anomaly Detection in Network Traffic

Anomaly detection in network traffic is critical for identifying potential security breaches, such as phishing, malware, or Denial of Service (DoS) attacks. Traditional signature-based Intrusion Detection Systems (IDS) have proven inadequate in identifying novel threats, prompting the adoption of machine learning (ML) and deep learning (DL) approaches. This study explores the integration of advanced ML and DL models with feature extraction techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) and Word2Vec to enhance anomaly detection capabilities. Models including Random Forest (RF), Support Vector Machines (SVM), Dense Neural Networks (DNN), and Long Short-Term Memory (LSTM) were applied to a real-world dataset of malicious URLs. The experiments demonstrate that deep learning models, particularly DNN and LSTM, out-perform traditional methods in detecting complex anomalies, achieving accuracies of 89.66% and 89.27%, respectively. This research provides insights into the application of hybrid ML and DL techniques for robust and scalable anomaly detection systems. The study also highlights the potential of future advancements in Transformer models and ensemble learning for improving real-time detection in dynamic and high-dimensional network environments.