Fraud prevention applies tailored rules and algorithms against a large and dynamic knowledge base, vetting inbound business transactions to provide real-time risk assessment and decisioning.
For banks, brokerage houses, and payment gateways, fraud prevention is vital. Both false positives and false negatives can have huge business implications. Failure in these areas results in losses to the bottom line and brand damage.
In about 750 milliseconds, a fraud prevention application has to review a user profile, test the request against patterns determined by data scientists, and make a “yes” or “no” decision as to whether the activity is fraudulent. Does this request look like fraud? Is it legitimate? A single request may result in a cascade of relevant database lookups. Latency is key to success here—the lower the latency, the more work can be done to test the veracity of each request. What makes this difficult is the large and ever-changing dataset of requester information. Was the card reported stolen? Is the charge over the limit? Is the transaction unusual? Was the request made online or in person? Is the device recognized?