is an extension of the company's FraudMAP Online product for desktop computers, which studies how a bank's customer uses an online banking site.

The software, which runs on a bank's servers, creates a mathematical model for how a person normally uses an online banking website and then looks for anomalies in their behavior, which could be a sign of fraud, said Terry Austin, Guardian Analytics' CEO.

The model for normal behavior can be created in about three weeks for a person who logs in two or three times a week, Austin said. FraudMAP Mobile looks for parameters such as what network a person is connected to, what device is used, the time when a person logs in and approximately where the person is located, he said.

Austin said the software generates fewer false positives than rule-based systems, which for example might cut off a person's access to their account if they log in from another country. "We are always looking at what you do as a user, so as a result we can be much more precise," he said.

FraudMAP Mobile will also spot some of the early preparation done by hackers to an account in order to make a fraudulent transfer go smoothly. For example, hackers might change a person's phone number in order to intercept one-time passcodes sent to a mobile device for transaction verification, Austin said.