Codetect Financial Fraud Detection With Anomaly Feature Detection

Abstract

Money laundering is one type of financial fraud that is well-known for diverting illegally obtained funds to
terrorism or other criminal activity. Complex networks of trade and financial transactions are involved in this
type of criminal activity, making it challenging to identify fraud firms and identify its characteristics.
Fortunately, the intricate networks of trade and financial transactions may be used to build the
trading/transaction network and the characteristics of the entities in the network. While features of entities are
descriptions of the entities, anomaly detection on features can reflect specifics of the fraud activities, the
trading/transaction network reveals the interaction between entities, making it possible to identify the entities
involved in the fraud activity. Hence, network, the majority of approaches currently in use only use one of the
two types of information, networks or characteristics. In this research, we offer CoDetect, a novel framework
for financial fraud detection that can use both network information and feature information. Additionally,
CoDetect is capable of detecting both the features associated with financial fraud activities and the fraud
actions themselves concurrently. Numerous tests using both artificial and real-world data show how efficient
and effective the suggested methodology is at stopping financial fraud, particularly money laundering.