Fraud is as old as humanity itself and can take an unlimited variety of forms. According to PwC’s 2020 Global Economic Crime and Fraud Survey (http://mng.bz/l2ny), 47% of global organizations have been the victim of a fraud (and many of the remaining 53% may have been without realizing it, so the number is almost certainly higher) for a total estimated loss of $42 billion. The European Central Bank reports (http://mng.bz/BK8J) that the total value of fraudulent card transactions annually amounts to €1.8 billion ($2 billion).
Fighting fraud, and more generally detecting anomalies in data, is a vital task that has enormous impact in multiple areas, such as finance, security, healthcare, and law enforcement. It has gained a lot of interest recently among machine learning practitioners. Whereas before, companies in most domains used a mix of human-based and fixed rules-based analysis, fraud detection is becoming more of an automated process, with machine learning playing a key role.