A reasoning agent for credit card fraud using the event calculus

We illustrate the design of an intelligent agent to aid a merchant to limit fraudulent payment card purchases over the internet. This is important because increasing fraud may limit the rise of electronic-commerce, and difficult because of the uncertainty in identifying and authenticating people remotely. The agent can advise the merchant what actions to take to reduce risk without complete knowledge of the circumstances. It can also negotiate flexibly to conclude transactions successfully that would otherwise be rejected. We use the event calculus to model the transaction system including the participants and their actions. The idea has applications in other distributed systems where incomplete knowledge of a system may be exploited by adversaries to their advantage.