AI glossary · Basic · Classic ML, recsys & search · Finance & insurance
Fraud detection
Fraud detection means automatically picking out suspicious, out-of-the-ordinary transactions such as card theft or scams among financial transactions like card payments and transfers, then blocking or verifying them. In the financial sector it is often called an FDS (Fraud Detection System).
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Industry AI
- Industries
- Finance & insurance
- Added
- 2026-10-10
How it works
When a transaction occurs, information such as amount, time, location, device and merchant type goes straight into the detection system. Rules such as 'payments in two distant locations within a short time' filter it first, and then a machine learning model compares it with the customer's usual patterns and assigns a risk score. Depending on the score, the transaction is approved as is, sent for additional authentication, or blocked. Flagged transactions are reviewed by analysts, and the results accumulate as 'fraud/not fraud' labels used to retrain the model.
Why it matters
Financial fraud such as card theft, account takeover and voice-phishing transfers causes direct losses for both customers and financial institutions. Fraudulent transactions finish within seconds, so they cannot be stopped by having people check each one, which makes automated detection essential. A well-built system reduces losses while letting legitimate customers pay without friction. That is why it is considered core infrastructure for financial institutions and is closely linked to anti-money laundering (AML).
Good to know
Fraudulent transactions are very rare among all transactions, creating a class imbalance problem that makes it hard for models to learn. There is also a balancing problem: too sensitive, and legitimate payments are often blocked, frustrating customers; too lenient, and fraud slips through. Fraudsters keep changing their methods to avoid detection, so data drift must be monitored and models updated regularly. Reasons for blocking often have to be explained to customers and regulators, so interpretability is also treated as important.
Example
Getting a 'Did you make this payment?' alert or call from your card company after making a large payment at an unfamiliar shop while traveling abroad is a typical example of fraud detection. The system judged the payment to differ from your usual pattern, gave it a high risk score and asked for extra confirmation. The same principle applies when a banking app briefly delays a transfer or asks for additional authentication when you send a large sum to an account you have never used before.