This paper presents a proposal for discovering anomalies in e-banking Web sessions by implementing different datamining techniques in a a graph-based environment. Online banking is a good example of how millions of costumers rely on virtual channels for business transactions. Nevertheless, due to multiple scandals regarding security flaws, it becomes complicated moving a business from a physical scenario to the digital world. Therefore, security applications become highly necessary. Monitoring systems like HIDS intend to create a more reliable scenario for companies but because of the number of sessions linked to e-banking Web servers it is barely impossible to detect fraud in real time. We propose a novel method for detecting anomalies in e-banking services by integrating efficient clustering systems based in sequence alignment and graph mining.