Forensic network analyzes intrusion evidence obtained to find out suspicious members and initiate step by step actions in an attack scenario. The evidence graph model serve as collected evidence. Depending on it one can form a framework that is based on hierarchical reasoning. Fuzzy inference comes in handy to comprehend host’s functional states from local observations. Graph structure analysis can be done through global reasoning to determine the potential attackers. We evaluate various techniques through obtrusion ferreting out datasets and trial and error results and establish that evidence graph model is compelling to detect multi-stage attacks. Then, for fraud ferret out problems, the data evolves continually from the system under consideration. Moreover, the underlying concept changes from time to time dynamically and is understood as concept drift. Mostly the frauds are rarely observed compared to the normal behavior of the system. It is very difficult or expensive to simulate fraudulent behavior from the system. Data mining warrants robust, dependable anomaly ferreting out systems. It is a fact that research so far happened has not focused much on graph-based data. Suppose that a real graph with weighted edges is known in advance and we are interested to find a method to classify vertices as strange? Answering this is quite important for applications such as: obtrusion ferreting out mechanisms while facing the fraud happening in credit/debit/calling cards and many others. We probe further on this here.