Artificially Intelligent Fraud Identification System - AIFIS
Vendo has developed a proprietary machine-learning algorithm based on neural network technology, that scores every transaction that goes through our platform. Further action is taken based on this score (i.e. Extra verification step with 3D Secure, Block the order, Send straight to manual review, etc.)
The AI takes into account more than 50 relevant criteria (ex. BIN per merchant, name per country, gender per site, the name contained in the email, random versus comprehensive password, customer behavior, returning customer, device recognition, etc). All of these criteria are scored against historical data and computed into a fraud score.
Real-time filters provide us with an extra layer of security for fraud prevention. Based on the continuous analysis, the risk manager adapts filters to the fast-moving fraud environment. Find below an explanation of the different kind of filters we use:
1. General Filters
The risk system chooses to block transactions based on elements such as:
- Payment method
- IP Address (Standard, Anonymizer, Satellite)
- Email domain
- Order price
- Password patterns
2. Velocity Filters
Velocity of count
Velocity is the rate at which orders are attempted. With velocity tests, Vendo detects suspicious transactions that arrive at a high rate and enforces distribution rules. Examples of tracking elements:
- Card Number
- Email Address
- IP Address
Velocity of change
Velocity of change detects an incremental change between order elements. Every change increases the probability of fraud and as a consequence the fraud score. Changes of any of the following combinations are tracked:
- Account Number
- First Name, Last Name
This check is provided by our partner gateways
Address verification is available for the US and part of the UK for card types VISA and MASTERCARD.
Manual Review of High-Risk Transactions
Risk Agents manually review transactions based on:
- Scoring from the AIFIS machine learning algorithm
- Alerts received from suspicious behavior
- Decision of the risk managers to focus on a particular affiliate, country, site, transaction type, etc.
- ACCEPT: keep transaction.
- REJECT: cancel, refund and blacklist account number, email and/or IP
- CONTACT SHOPPER: contact shopper to verify identity. No answer from the consumer leads to a refund and blacklist of the account
Third-party fraud alert providers
Vendo implemented 3rd party fraud alert solutions that have been valuable for our merchants. One such service provides alerts from issuing banks from unrecognized transactions, which we can handle in order to avoid a chargeback.
Vendo generates a Fingerprint ID for each device. This allows us to track the device purchase history as well as to detect unusual behavior attached to each device.
Botnet and proxy detection
Vendo has implemented botnet and proxy detection systems to stop fraud. The system identifies patterns and declines the transactions before sending them to the bank.
Routing Rules and Specific Fraud Prevention Solutions
Vendo can use verification steps during the purchase process to improve the routing and scoring of a transaction. Some of these features include:
Prevents Bots from accessing the system and discourages fraudsters. Moderate to high-risk transactions are forced to go through a reCAPTCHA verification process. Passing the verification often proves a real person is behind the transaction.
High-risk transactions are submitted to a phone PIN verification procedure: end-users must validate a code sent to their mobile phone numbers to initiate the transaction.
Updated almost 3 years ago