ProbayesCard

Probability theory helps fighting against fraud

ProbayesCard is a Fraud Detection solution for Debit and Credit Card payment transactions and withdrawals.

ProbayesCard can:
 detect fraudulent transactions in real-time and a posteriori,
 generate transaction alarms in real-time,
 generate card alarms, ordered by fraud probability,
 compute the effectiveness (performance and relevance) of the models,
 update its own parameters from recent data.

ProbayesCard returns a fraud probability and an advice on how to handle the transaction. This advice depends on the probability threshold given by the risk manager. This allows to:
 generate an immediate alarm,
 return a "referral",
 decline the transaction, and stop the fraud.

ProbayesCard computes a unique metascore, wich is the fraud probability using the answers of:

• Bayesian models designed by Probayes to fight different fraud schemes, based on the discrimination of fraudulent and legitimate profiles:
  • Cross border Counterfeit Card,
  • Card not present,
  • Stolen card...
• External scores (VISOR, Financial Institution’s tools, expertly based rules...).

ProbayesCard takes profit of customer’s existing tools and enhances them.

If needed, other models can be designed by Probayes and added to the system.

Each Bayesian model fits to the bank data and adapts itself to the evolutions of legitime cardholders and defrauders thanks to a probabilistic learning on payment/withdrawal authorizations and fraudulous hictoric files.

Parameters can be customized by the risk manager. He can test and validate models and their parameters on test and historic files. The following indicators are provided for different thresholds:
 performance: maximize the fraud coverage (fraud detected/fraud)
 relevance: minimize the false positives (actual fraud/alarms )
 number of alarms
 fraud amount