Robust Anomaly Detection in Network Traffic: Evaluating Machine Learning Models on CICIDS2017