Fraud Prevention Engine - Beyond Cyber Security

InsightLake built a robust Cyber Security solution, which allows organizations to detect intrusion, anomaly detection and internal activities closely. Its Fraud solution assists organization's to avoid risks/fraud in real time easily.


Organizations lose billions of dollars around the world because of fraud. Fraudsters use variety of methods to commit fraud. Fraudsters use fake identity to apply for new accounts and perform transactions. They steal identities using various methods. Please check another InsightLake blog post on Credit card fraud, which lists various activities fraudsters perform.

Fraud is classified in two classes.
1 First party fraud - Customers commit fraud by using fake identity/information to open an account, making transactions without any intention of paying back. They defraud companies either by busting out or disputing.

2 Third party fraud - Hackers or other fraudsters steal companies customers accounts and perform transactions.

There are different categories of fraud:
1 Application fraud - New application with wrong identity or information

2 Transaction fraud - Transactions through various channels like online, mobile, POS, check etc.

Insight Lake Real Time Decision Engine

InsightLake platform has built in real time decision solution, which includes machine learning models for different use cases, scoring and scenario analysis models, extensible business rules framework. This engine power our key solutions like:

  • Cyber security detection and remediation
  • Real time advertising bidding/placements
  • Fraud prevention

Fraud Prevention
Authorization of credit card or direct payments for online retailing companies, bricks & clicks, e-businesses, and traditional brick & mortar have always been prone to the risk of fraud. Our fraud detection & management solution helps debit & credit card companies, gateways, and payment processors in online fraud detection, management, and prevention.

Realtime - decisioning in millisecond

To detect transaction fraud, InsightLake's real time decisioning module allows parallel business rules execution on transaction information elements, scoring and machine learning model based anomaly detection in real time.

Detection in milliseconds enables organizations to perform remediation actions like stopping the transaction, alerting or incidence/case creation.

Machine learning platform

Fraud management solution utilizes many optimized machine learning models for scoring, anomaly detection etc.

Flexible feature engineering enables customization of models very easy. Running multiple models in parallel and feedback loop allows model promotion based on performance.


Scoring is the core part of fraud detection logic, assisted with rules it provides a valuable metric based on which a decision can be taken.

Scoring model allows complex formulas with configurable weighted parameters.

Scenario analysis enables weights to be modeled appropriately.

Anomaly Detection

Anomalies or outliers are the most important indicators of big change of pattern, which could be from a fraud scenario.

Fraud management solution allows anomaly detection with seasonality to allow real outlier detection.

Transaction Fraud

Transaction fraud is the main contributor of fraud losses. This type of fraud can be detected with Fraud management solution's robust engine, which includes scoring and business rules to detect fraudulent authorization or transaction with low number of false positives.

Application Fraud

Fraud management solution provides following features to detect application fraud:

  • Machine learning models to detect application information fraud like income groups, age, geo etc.
  • Fraud network detection using fuzzy matching
  • Business rules allowing criteria based screening

Pattern Matching

Other than fraud decisioning InsightLake's Fraud solution allows complex pattern matching and alerting in real time.

Fraud operations team can build pattern models and apply them in real time or explore those on historical data.


Fraud solution can be easily integrated with organization's existing network. It could be put in the path to allow/deny or be deployed in parallel to do active monitoring, alerting and scoring.

Easy integration with:

  • Alerting/Operational dashboards
  • Case management/Incident response
  • Existing Hadoop Cluster

Hadoop Based Scalable Solution

Hadoop based Fraud solution works for small or large companies and could be scaled to handle from few hundreds to billions of transactions per day. Companies can start from small deployments and add more instances to scale easily.