Every year Bank Info Security releases their comprehensive Faces of Fraud Survey that looks at the biggest trends and current state of bank fraud. The most alarming statistic from the 2014 results is that 62% of companies indicated that they first learn of fraud incidents only after their customers notify them. (Tweetable)
Here are some other insights from the survey:
- 36% point to the difficulty of integrating data from various sources as one of their biggest challenges
- Only 11% are using big data analytics in fraud prevention
- 65% say payment card fraud is the most common form of fraud they see
Considering that the banking industry spends Millions of dollars on some combination of controls to detect fraud, including security cameras, fraud monitoring systems, Positive Pay and debit blocks, it’s pretty disheartening to know that customers are among the best fraud detectors.
What if in 2015 we flipped that percentage? If 62% of banking fraud was detected before the customer even notices, the impact of fraud would significantly decrease in terms of money lost, employee time spent and damage to the bank’s reputation. But why stop there? Ideally we would completely remove the customer from the fraud detection equation.
The Problem With Current Fraud Solutions
One of the biggest challenge that banks face is the sheer volume and the variety of data that needs to be mined and coordinated in order to detect fraud. Banks don’t only process ATM and Teller data: there’s online banking, drive-thrus, surveillance cameras, sensors, access control, and more.
Current fraud detection and prevention solutions are supposed to mine this data and create alerts whenever they detect potential fraud. But it is pointed out by the survey that many of these solutions don’t truly integrate different types of data. They just combine fraud alerts together from several channels, resulting in three major problems:
- Thousands of potential fraud cases are created without high accuracy
- Many of these end up being false positives
- Without data integration or the ability to review thousands of cases, some fraud goes undetected
The first two problems make it impossible for investigators to manually review each case in a timely manner. If potential fraud was flagged in the transactional data, the investigator would also need to correlate it with the video evidence and any other data that corresponds to that transaction. This can draw out the investigation process by hours, if not days, more than enough time for the criminal to commit a second ATM fraud.
Properly integrated and managed data is a critical element of fraud detection because it reveals suspicious and unique patterns which would otherwise go unnoticed. Consider a situation in which the client is using the ATM. The video footage looks normal and the ATM system detects nothing irregular. However, when integrated on a central platform, the system recognizes that the client has been standing in front of the ATM for 5 minutes and has yet to make a transaction. This would trigger an alert to offer assistance to the client, or to ensure that the machine has not been tampered with.
How Big Data and Contextual Analytics Can Change This
Contextual Analytics holds tremendous promise for preventing and detecting fraud in the banking sector. By layering different data types , this technology has the ability to identify suspicious events and anomalies hidden within a bank’s ever-growing stores of big data.
“But right now, only a handful of banking institutions throughout the world are adequately using big data to enhance security” says Dr. Anton Chuvakin, security expert and analyst for the consulting firm Gartner. (source) As mentioned, only 11% of the survey respondents indicated that they are using big data analytics for fraud prevention.
Although it is often overlooked in favour of transactional data, video can be an incredibly valuable source of intelligence for financial institutions: as shown here and here, recorded video is the biggest big data, as well as the most context-rich. As the name suggests, we developed Contextual Analytics as a way to use video in combination with all other data sources to reveal the unique context around any event or transaction and to improve fraud detection accuracy and speed.
Here is an example of what bank fraud detection looks like when leveraging Contextual Analytics,
- A man takes out $475 at an ATM drive-thru. The event details such as date, time, location, card, and license plate are all recorded and linked to video footage of the transaction. He then makes a second withdrawal using a different card, and then a third. The system recognizes multiple cards being used by a person with the same licence plate, this triggers an alert in the system. Within seconds, an investigator can be notified and begin to review the case with all the the required information available in one place.
Although we generated the data shown in the image for the purpose of this post (hence why we blurred the man's face), this type of situation is very real and detecting it early means catching the fraudster faster.
Luckily, more and more people are talking about big data analytics in the financial sector. As traditional tools are proving insufficient, we believe it's time to increase automation, leverage a bigger variety of data and demote customers from their role as fraud detectors. (Tweetable)