By Vinay Samuel, Founder and CEO, Zetaris
We all know that access to high quality data at the right time is critical for making the right decisions and the right time. But a lack of access to high quality, timely data doesn’t just increase the risk of a bad business decision. It also increases the risk of fraud.
The challenge with fraud is that reacting after the event doesn’t often reduce the losses. We see this frequently in the finance industry. Once a criminal or threat actor compromises access to funds, they can transfer and move their bounty so quickly that recovery is almost impossible. The power and convenience we covet as customers has created a lucrative getaway vehicle for 21st century bank robbers.
The key to detecting fraud before costs escalate is through real-time access to data and monitoring for anomalous activity right across your data environment. That includes across networks, within applications and user behaviour.
Organisations trying to minimise the risk of fraud need to ensure they have fast access to high quality data. However, traditional approaches to bringing together disparate sources of data have never been able to fully meet this challenge. While they have been able to apply some degree of monitoring and analytics to operational data, pulling together the different pieces has relied on extracting data, transforming it and loading it into a centralised data store for analysis.
This process, often called ETL (Extract, Transform, Load), can be extremely complex and doesn’t support simultaneous, real-time analysis of multiple data sources. What’s needed, in today’s world, with data that is constantly changing from more sources than ever before, is a way to analyse data from disparate sources so fraud can be detected faster. This is why a data mesh approach is critical.
With a data mesh, data is freed from the databases in which it’s stored. Data owners determine how to share data and make it available to authorised parties. Rather than needing to copy data to a centralised data warehouse or data lake, it becomes possible to query data without extracting it. In fraud detection, being able to monitor across disparate data sources simultaneously, means faster response times to thwart malicious activity. Even saving a few seconds can be the difference between avoiding a loss or a thief stealing data or funds and removing it from your reach.
Detecting fraud is complex and requires multiple inputs to answer the question about whether an action is a sign of fraud taking place or an indicator that fraud is imminent. For example, being able to detect that an employee is logging into a system from an unusual location or at an unexpected time could be a sign. Similarly, a customer making an unusually large transfer could be an indicator. But it may be more subtle.
This is why machine learning and AI are so critical in fraud detection and prevention. It has the ability to make connections between data sources that are changing faster than people can process. But those models require access to fast and accurate data which is why a data mesh that allows queries to be executed against live data is so important. Data from more sources can be looked at faster to uncover connections and detect abnormal behaviours rather than focussing on just a limited number of accounts or activities.
The key to fraud detection and prevention is fast access to reliable data. By employing modern approaches to data access, such as the data mesh and data fabric, it’s possible to detect the signs of fraud faster and prevent losses. Today’s cybercriminals are extremely sophisticated. And while they can be very stealthy, they always leave some evidence of their activity. Being able to detect that evidence faster can be the difference between quickly stopping a criminal in their tracks or a substantial, perhaps business-ending, loss.