1 What is fraud and fraud detection?

 

This chapter covers

  • What is fraud? How does it affect us? How do we detect and stop it?
  • What is machine learning and how does it help fight fraud?
  • What is deep learning and how can it detect the most complex types of fraud?
  • Exploring some of the different types of online fraud.
  • How do we evaluate a fraud detection system?

Welcome reader, to the world of fighting fraud. Fraud is at least as old as 300 BC - when a Greek sea merchant named Hegestratos tried to run off with a ship-and-cargo protection loan. He took a transport insurance policy where he got a sum of money equal to the monetary value of his ship and cargo. He would have had to repay the money with interest upon the safe arrival of the ship and cargo to the destination. The cargo, in his case, was corn. In an attempt at fraud, he planned to sell the corn, sink the empty ship, and therefore keep the loan money. He was caught in the act and drowned while trying to escape. Ever since, fraud is everywhere - more today than ever before - in the form of fraudulent card payments, fake social network bots, digitally forged documents, email scams, and so on. In 2020, Juniper research suggested that online payment fraud alone will incur losses worth over $200 billion over the next five years - more than the GDP of many countries (www.juniperresearch.com/press/online-payment-fraud-losses-to-exceed-200-billion).

1.1 What is fraud?

1.2 How do we detect and prevent fraud?

1.2.1       Rule-based fraud detection

1.2.2       How far can rules take us?

1.3 Enter machine learning, exit fraud

1.4 Deep learning to fight them all

1.5 Types of fraud

1.5.1 Phishing - email phishing, smishing, vishing

1.5.2 Identity fraud

1.5.3 Online transactions fraud

1.5.4 Digital document forgery

1.6 Evaluating fraud-detection systems

1.6.1 Accuracy

1.6.2 Confusion Matrix

1.6.3 Precision, recall, and f-score

1.6.4 Precision-recall curve

1.7 Summary