Up to 5% of global GDP is estimated to be lost through fraud and corruption, costing the global economy around $2.6 trillion annually. Fraud investigations are highly complex and labour-intensive; they must also be extremely rigorous, and this can generate a high rate of “false positives” for in-depth investigation.
Anomaly detection deals with the problem of finding patterns in data that depart significantly from the expected behaviour. Being able to detect such anomalies is crucial in domains such as fraud detection for credit cards and bank transactions, insurance claims, and money-laundering. Being able to detect such anomalies in a timely manner allows one to take actions and restrict or limit the negative consequences of such anomalies. For example, anomalies in financial transactions can indicate credit card theft or money laundering operations, which can be stopped if acted upon swiftly.
This project will carry out a proof of principle study for detecting anomalies in networks, based on a similarity matrix between networks.