Fraud Detection

Fraud detection systems score transactions, accounts, or claims for abuse using rules and machine learning.

Software that flags suspicious money or identity activity for review.

Fraud detection systems score events using supervised and unsupervised ML, graph features, and rules. Adversaries adapt, so monitoring and retraining matter.

Fraud is a primary production ML domain in finance and commerce with clear loss metrics.

• Payment fraud
• Account takeover
• Claims fraud screening

• Reduces loss when tuned well
• Prioritizes investigator workload

• False positives harm customers
• Concept drift from new attack patterns

• A single model score is not a complete fraud program

Related sectors?

Often Finance or Insurance depending on the buyer.

• NIST AI RMF

• Imbalanced learning and graph-based fraud detection literature