Detecting Fraudulent Communities in Financial Networks Using Hybrid Classification and Ranking Approaches

Suspicious communities refer to networks or organizations that display untypical behaviors in different fields such as in cyber security, social networks, and finance. These groups are also identified by the peculiar patterns of communication, abnormal transactions rates, and links to established criminal aspects. The well-coordinated and deviant characteristics of the members, including inconsistent timing and amount of interaction are often the indicators of possible fraudulent plots or money-laundering. It is important to identify these communities to detect and intervene on the illegal activities at an early stage. The paper presents a model of the detection and identification of the suspicious communities engaging in money-laundering transactions. The proposed framework identifies the highly suspicious nodes through a classification layer first and then identifies the suspicious communities around the suspicious nodes using three different algorithms. The algorithms have various strategies of ranking and classification to identify suspicious communities. The suggested framework was tested on two banking datasets and it was found to be able to find fraudulent communities with a high level of success.