There are critical characteristics of fraud, such as, continuous behavior changes, exponential growth rates, unpredictability, sophisticated advanced fraud techniques and others, that when addressed by the proper technology, can be successfully curtailed. However, the present generations of fraud solutions are based on methods that do not efficiently respond to these characteristics and certainly cannot respond to the constant changes in the fraud world.
Most fraud cases, if they are not detected before the crime is committed, will become a loss.
Criminals today are continuously adapting their techniques as well as concocting intricate fraud schemes. Without new aggressive prevention procedures in place, fraudulent transactions can thrive indefinitely.
Current Generation of Fraud Product Ineffective
Detecting and preventing fraud is a dynamic and continuously evolving process. Therefore, a static set of “Neural Network Signatures”, "Data Mining Decision Tree" or “Rules” has only limited value.
Today’s available solutions are unsuited for preventing fraud. Why? Because current technologies must be aware of fraud methods before these prevention efforts can be implemented, so they are unable to detect new types of fraud as they occur. Furthermore, they require massive amounts of historical data to recognize patterns.
These technologies lack adaptation to new types of fraud.
Like current anti-virus applications, they are outdated as soon as they are released.
Today’s technologies are clearly inadequate in a world of ever-more clever thieves.
Dynamic Self-Adaptation maintains continuous accuracy
iPrevent, Brighterion’s Fraud prevention product is the result of 500 man-years of research and development. In contrast to current static anti-fraud products, Brighterion’s iPrevent solution is uniquely self-adaptive. The comprehensive intelligence built into iPrevent allows the customer’s model to gather intelligence and incrementally learn as transactions occur.
First Generation of Fraud Detection
Scientists initially created programs called Fraud Scanners that detected known cases of fraud by using signatures that identified instances of fraud. These signatures were classified and stored in a database. The fraud scanners compare each transaction to the “known” fraudulent signatures stored in the database. If a match occurred, the transaction was classified as fraudulent.
This method is ineffective and requires updating the database as soon as new fraud activity is discovered. With new fraud cases arising daily, a company needs to update their computer’s anti-fraud database continuously. Furthermore, unless an employee is forewarned about a fraud case, the fraud activity has already become a cost factor to the company. This method is labor intensive and as a result, fraud detection is low.
Current Tools
The current leading fraud detection methods still use Neural Network algorithms, (for more information, refer to the excellent article from Gartner “Online Fraud Prevention White Paper for the E-commerce Fraud Prevention Network, Source: Gartner Consulting, Adam Hils, 02/12/2001”) A Neural Network learns relationships between sets of input and output patterns, and can then intelligently respond to new inputs using the experience it gained during training.
Neural Networks utilizing supervised learning techniques, such as Back Propagation networks, have proven themselves in several practical applications, such as classification, generalization and prediction. However, having to know the desired output for each input before training begins is often a prohibiting limitation. When the desired outputs for the training input patterns are unknown, the fraud situation is a perfect example of how new incidences cannot be detected in real-time. Hence, there is a crucial time lag between detection and prevention.
Neural Networks, Data Mining, statistical modeling or expert systems (rules), have been applied to fraud detection. For such methods to be effective there must be a large database of detected cases (signatures) and the characteristics of fraud must not change. As mentioned before, fraud does change and evolve rapidly. Therefore, these tools are not effective because criminal creates new schemes and adapts new behaviors.
Brighterion's Value Proposition
“Catching Previously Unknown Fraud = Preventing rather than Detecting”
Brighterion is a pioneer in the development of Smart-Agents technology and is the first company to apply it to a commercial anti-fraud solution. In Smart-Agents technology, by contrast to supervised models where the system needs to be trained with good samples, multiple independent Smart-Agents are each assigned one or more goals, the Smart-Agents then interact and negotiate with each other in order to reach their individual and collective goals.
Our flagship fraud product, iPrevent, is uniquely capable of identifying “any abnormal behavior” and therefore, the only product on the market with the ability to identify and stop the most elusive and costly “first fraud” (previously unknown fraud).
Brighterion’s patented technology suite includes ten artificial intelligence and advanced technologies that work in unison to solve complex problems and to improve accuracy. iPrevent analyzes transactions, using a combination of these technologies, to deliver a collaborative result that is significantly more accurate and unparalleled in the industry.
iPrevent has clear market advantages:
- Model and predict behavior with greater accuracy
- Customize the analysis model to unique requirements
- Identify suspicious behavior without extensive profiles or historical data
- Self-adaptive and learning fraud model
Custom models equals more accurate results
Custom modeling is currently a highly manual, costly process and typically restricted to the largest business enterprises. The vast majority of anti-fraud customers use products and services that are generic and without sensitivity to an individual enterprise’s industry, products, customers, or fraud attacks.
Brighterion has automated the modeling process to such a degree as to make custom modeling require significantly less expertise and is faster than traditional non-specific business alternatives. iPrevent provides customers the dual benefits of highly accurate fraud prevention : less fraud and lower false positives.
How It Works
The human immune system produces antibodies to protect the body from known and unknown ailments; iPrevent for Fraud produces “antifraud counter measures” to protect your business from known and unknown fraudulent activities. The analogy between the human immune system and iPrevent illustrates the software’s functionality. Human immune systems protect the body from many known ailments. The immune system automatically adapts to attack new and unknown dangers to the body by spewing out millions of white blood cells. Most importantly, it creates antibodies to protect the body with knowledge from previous illnesses. This is the framework on how iPrevent operates.
The Power of 10
Brighterion brings to the market the first fraud solution that makes effective changes during run-time, to adapt to new and previously unknown fraud behaviors and conditions. iPrevent achieves this by using Brighterion’s Smart-Agents technologies which can recognize unknown types of frauds in conjunction with 9 others advanced technologies : Neural Networks (Brighterion has a patented algorithm, as well as another 12 neural network algorithms), Case Based Reasoning, Constraints Programming or Optimization tools, Statistic tools, Business Rules, Genetic algorithms, Velocity Analyzer, Fuzzy Logic and Data mining . These technologies are fully integrated with one another within an application to provide optimal, next generation fraud prevention solution.
Stop Identity Theft and account takeovers
Identity theft is a serious problem for all financial institutions. A report issued in September 2003 by the Federal Trade Commission estimates that more than 9 million Americans were victims of some type of identity theft within the previous year.
Identity theft occurs when private information gets into the wrong hands. It may be available through public records and databases, medical claims forms, or by illegal access to commercial databases. A criminal can use this sensitive information to open a new account in another name or take over an existing account.
Why identity theft occurs
1. Creditors assume that the person applying for a new account is, in fact, that person.
2. Current fraud detection systems are separate, product-specific, and isolated.
Brighterion’s Solution
Any solution that relies on keeping information out of the hands of thieves will be inadequate. iPrevent stops identity theft by detecting atypical events at the earliest stages ; before the damage is done. Brighterion solution consist of two majors modules:
iPrevent Automated Decision
iPrevent Automated Decision (IAD) is an authentication system that focuses on determining the validity of personal identifiers. With IAD’s three levels of risk management (validation, verification, and authentication), financial institutions can expedite customer acquisition and validate large numbers of applications in real time.
iPrevent Cooperative Abnormal Behavior Detection
Indentity thieves often exploit the lack of colloboration of fraud systems running in different business units (Check, ATM , Credit Card, etc.). iPrevent solves this problem by merging and unifying in realtime the results from the different iPrevent instances. This collaboration will detect atypical events, stopping identity theft before the damage is done.
iPrevent Implementation
Implementing iPrevent consists of six steps: creating the custom model, installing the model on the network, processing transactions, managing output, reporting and learning.
Implementing iPrevent consists of six steps: creating the custom model, installing the model on the network, processing transactions, managing output, reporting and learning.
Customize Your Model to Your Unique Environment
With iPrevent, a model can be created automatically with no need for any programming, IT involvement or artificial intelligence knowledge. Brighterion's development team spent more than 5 years making the design of models absolutely transparent and easy to use.
In "Automatic Mode", a highly intuitive wizard-based interface and templates enable you to easily create a customized model from any relational database: the neurons, the cases, the Smart-Agent, the fuzzy set and the data mining parameters that are best adapted to your business data. iPrevent will automatically extracts from your historical data the fraudulent behaviors : the Smart-Agents, the decision tree, rules, cases, and pattern signatures to best protect your business.
The "Expert Mode" enables modelers to enhance the models with iPrevent Busines Rules, iPrevent Velocity Rules, iPrevent Flexible/Fuzzy Logic, iPrevent Constraints and iPrevent Genetic Algorithms.
Installing the Model on the Network
iPrevent is based on non intrusive and flexible Architecture that allows you to deploy it in hours, leveraging your existing enterprise infrastructures. The iPrevent Fraud Model is incorporated into the customer’s transaction processing as a network service. Multiple instances of the model can be installed and used simultaneously. This means iPrevent can be set up to handle different types of fraud through out your organization at the same time.
Processing Transactions
iPrevent interface with multiple platforms using industries standard protocols : TCP (ISO8583 Format, TCP Intterface, Various API..) , HTTPS, XML, SOAP / Web Service. This reduces development and deployment costs.
iPrevent can handle up to 500 transactions per second per server (entry level hardware). iPrevent can be deployed in many servers.
Managing Output
In addition to making the allow/review/deny decision in real-time, iPrevent includes Live Log for output management. Live Log is a thin client used by monitoring or investigative personnel that displays each of the records that have been referred to that staff member. These records include the detailed results of the model, including the reason for finding the transaction to be suspicious or fraudulent.
Reporting and Case Management
1. Reporting Tools: automatically deliver various reports to any destination on the network.
2. Case Management System: Browser based case management system to enable investigation & decision making. An investigator can :
- Open/close/ release
- Add comments, transfer a case
- Update the status and the priority of a case
- Alert escalation : Supervisor notification/ email
- Manage fraud cases by :
- Suspect orders
- Users
- Servers
- Customers
- Merchants
- Organizations
- any other field
- Assign a resolution to the case and an assessment to each transaction
- And more.
iPrevent Monitor Tool
-
Controls the different iPrevent servers
- Start, stop
- Pause, resume
- Enable/disable statistics
-
Displays the statistics related to the servers activity
- For each type of message, number of messages received
- Average, minimum and maximum response time for the different messages.
- All the models are automatically encrypted
-
All data exchanges are encrypted,
- During the transactions processing
- Between the web server and the DMS during all transfers of data
- Only authorized users can have access to the models and the production data.
Maintaining accuracy is key to iPrevent’s continuing effectiveness. iPrevent is self-adaptive. The Incremental Learning enhance the intelligence of the models automatically with the F+/F- transactions.
iPrevent is used daily by many financial institutions around the world.