For the past 20 years, web tracking has raised worries among privacy advocates and authorities responsible for data protection. Researchers have proposed several machine learning-driven remedies to identify Web trackers in an automated manner. While those have displayed potential, they have primarily remained as proofs-of-concept. This work extends on t.ex-Graph outlined in our previous work [36]. The aim of this model is to distinguish benign from tracking hosts by considering their centrality in the network, and data flows to them. Based on the results of our previous work, we abandoned the SLD-based approach. Consequently, we made slight modifications to the feature vector. Our classifier’s performance is comparable to its original version, and we tested its cross-browser and longitudinal performance. Our results indicate that while the cross-browser performance significantly decreases, the longitudinal performance maintains a high level.