As digitization continues to accelerate, the volume of financial crime, and online financial crime in particular, will also increase. Financial crime 4.0, a term that GBG, a global digital identity expert and technology specialist in fraud and compliance management, identity data tracking and intelligence, invented to encapsulate new models of financial crime, will continue to s ‘exacerbate.
Financial institutions would not only need to detect fraud, but also manage the accuracy and efficiency of fraud detection. Fraud detection solutions that achieve lower false positive rates will be key to helping banks minimize the number of fraud alert investigations and analyzes.
While traditional rule-based detection systems are moderately effective, there are two main gaps in detection accuracy. The rules don’t catch enough fraud, and they produce too many false positives.
In this area, machine learning can play a role in identifying and mapping new patterns, tracking highly suspicious activity and known frauds better than a traditional rules-based engine.
Dev Dhiman, Managing Director for Asia-Pacific at GBG, says machine learning helps financial institutions improve two fundamental processes: customer experience management and financial crime prevention.
âReducing false positive rates isn’t just about streamlining internal processes and costs, it’s also about enabling legitimate customers to onboard and deal successfully. At the same time, with the entrenchment and escalation of Financial Crime 4.0, a rules-based system would ultimately be insufficient to detect and track emerging fraud typologies, âsaid Dhiman.
GBG Machine learning uses algorithms to uncover inconsistencies in behaviors and activities, analyze them, and report anomalies to historical sets of personal data observed within the company or across the industry.
The feature engineering component in GBG Machine learning creates the predictive analytics capability for the model. It allows the machine to constantly create accounts and customer profiles, the typical behavior of that customer, and constantly compare new behaviors to typical behaviors. The anomalies are modified to help the model detect and isolate possible emerging and new fraud patterns.
For example, if an account is hacked using data from breaches, scams, or phishing activity, GBG Machine learning can isolate inconsistencies in the behavior of the fraudster. If the fraudster started making daily transfers just below the daily limit, which is atypical behavior of the original customer, GBG Machine Learning is able to detect this anomaly and trigger an alert.
GBG Machine learning creates models that are sequenced to be activated with triggered rules or without triggered rules. By doing this, GBG is able to increase the accuracy in identifying false positives by focusing on the biggest problem for that client.
Unlike traditional rule-based systems that flag potential fraud because a series of rule conditions are not met, GBGThe goal of using machine learning is to find invisible and implicit correlations in real-time data and automate the detection of new possible fraud scenarios as quickly as possible, and determine if the anomaly is due to unusual or suspicious activity, rather than unintentional human error.
Adam Emslie, Head of Analysis at GBG, says the most important factor about machine learning models is not that they are feasible, but that they meet the standards of efficiency and accuracy demanded by customers and industry.
âFrom a unique end-to-end solution, GBGDigital Risk Management’s digital risk management and intelligence platform feeds data from origin and transactions directly to machine learning models, avoiding the need for data mining, enabling training and faster model rollouts than the industry typically offers, âhe said.
Fraud and compliance goals can be achieved through multiple workflows with the GBG Information Center. Fraud analysts are able to prioritize and improve the accuracy of detecting and preventing larger fraud challenges.
As fraud cases continue to increase on digital channels, GBG incorporated device-level information as a feature, which adds a significant improvement to the existing feature set.
This means GBGThe machine learning model ingests fraud, known historical data and behavior, as well as data received through GBG Information Center, which includes IP, address, identity, e-mail, social networks, phone /SIM, device behavior, and any other third-party data from a fraud consortium or third-party vendors.
Look ahead, GBG plans to add feature profiling and bivariate analysis reports for feature selection. With feature profiling, GBG analyzes which fields are really useful and can add value to fraud detection without cluttering the model.
Bivariate analysis will allow GBG to analyze the fields of value and discern which ones have a higher fraud rate.
GBG also provides more information to explain the scores produced by the models. It already returns the overall importance of functionality – the importance of functionality ranked equally across all customers and applications – to fraud reviewers. It aims to add the importance of local characteristics – specific variables / characteristics that contribute to a particular customer application and a fraud score, categorized by local context – to better understand why the model produced a particular score.
GBG also provides information on any potential bias against certain segments of the population due to the release of the model.