Are you who you say you are?

Big data analytics can be harnessed to verify online identity

The explosion in online e-commerce has been paralleled by a corresponding growth in online fraud.

Confidence tricksters who once put their minds to creating the perfect sting are now using a new range of online tools to appropriate identities, hack databases and systems, and appropriate other people's funds.

At the same time, the rise of social media has created a new world of trolls and fake identities, some relatively benign, others malicious in intent and action.

According to Felix Tan, a lecturer in the school of information systems, technology and management at UNSW Business School, fraudsters are having some success in this technology race.

"The same type of technologies the banks and IT companies can use to overcome fraud are accessible to the fraudsters themselves, so in some cases, the fraudsters have shown that, for now anyway, they are a step ahead," Tan says.

'Unlike traditional fraud detection methods that are unable to apply large-scale data analysis, business analytics can provide automated real-time fraud detection'

FELIX TAN, DANIEL CHEUNG & ZIXIU GUO

Validating the individual

Estimates of the extent and cost of online fraud vary wildly. Internet payments security firm Cybersource estimates that $US20 billion was lost to fraud globally in 2012, with the figure growing at twice the rate of online commerce.

The estimate from the Association of Certified Fraud Examiners is that global fraud cost is more than US$3.5 trillion of gross world product.

Tan, however, believes the cost of fraud is too difficult to accurately quantify.

"You could say that it is a billion dollar loss to Australians but what about the social and psychological loss, and all the therapy and remedial work that needs to be done over time?" he says.

Automated fraud detection

There have been some innovative technological responses to online fraud, and one of them is the subject of a recent case study by Tan and UNSW Business School colleagues Daniel Cheung and Zixiu Guo.

The three academics have been looking at the "data fingerprinting" solution developed by Irish vendor Trustev, which harnesses the power of big data to deliver real-time online identity verification.

"With traditional fingerprinting, the impact of skin friction ridges could be used to match personal identity but the concept [has been] extended to online identity verification," the authors say.

"As a result, data fingerprinting becomes a means to combine the elements of a digital identity with dynamic verification identity to formulate a score for fraud. This score then enables customers to make decisions driven by facts and data."

Instead of validating a credit card number, for example, Trustev focuses on validating the individual making the transaction, not just the payment method. Trustev's approach is to make sure that the person making the payment, or posting a comment, is who they say they are.

To achieve this involves a combination of transactional, behavioural and social data which is integrated and processed by an algorithm to create a score, which forms the basis for the decision on whether to approve a transaction or identity.

The researchers have identified 11 key components of the data fingerprinting solution, from algorithm analysis to data management and machine learning. These create more than 60 sub-score calculations which are then tied in to the full Trustev score, which is calculated out of 100 and delivered to the merchant or user.

"The Trustev case study is significant, because the unique thing about Trustev is that they have been able to harness data analytics to empower customers to make the right kind of decisions," says Tan. 

"This is the first case study we have encountered that combines data analytics, algorithmic principles and machine learning into a solution. It's something we have not yet seen, so there is a level of sophistication that puts it ahead of the curve.

"It also has the power to harness social media as well, and we have never seen a company monitor social data, capture and encode it and integrate it into an algorithm to make a real-time decision engine."

Trustev has particular application in blocking trolls on news websites, blogs, or on gaming sites. In most  present situations, people who have been blocked return – hydra-like – with a new identify that cannot be traced to any previous incarnations. Trustev integrates with a site's back-end system to automatically identify and ban these people.

It also has e-commerce applications. US electronics retailer RadioShack deployed Trustev across all of its stores, and online, at point of sale and check-outs. When customers made an in-store purchase, for example, Trustev would simultaneously scan more than 260 online data points to ensure the buyer was legitimate and confirm their identity, and do this within 0.25 seconds.

The result is that real users, who may have been excluded under a traditional rule-based approach, are unblocked. Using Trustev, one corporate customer increased net revenue by 11% within four weeks of implementation, while another saved 300,000 euros a month.

'Mobile technology, the cloud, ERP – all these buzzwords are going to be synced with big data to produce the next platforms for innovation'

FELIX TAN

Next platforms for innovation

Trustev's work is not, however, the first time big data has been used in fraud detection. The authors note the example of South African insurance company Santam, which implemented an analysis tool from a leading IT vendor which integrated "unstructured datasets into a single analysis framework".

The solution improved fraud detection and sped up claims processing, saving US$2.4 million on fraudulent claims in its first four months.

The practice, say the authors, "is one of the most promising features of business analytics, for detecting and reducing the impact of fraud."

"Unlike traditional fraud detection methods that are unable to apply large-scale data analysis, business analytics can provide automated real-time fraud detection," they write.

Tan believes the Trustev case study shows the potential to use big data.

"Big data creates the opportunity for enterprises and it can deliver fresh approaches to tackling age-old problems," he says.

"This is the only solution we have studied thus far that has the ability to capture more data than we have ever done before, and not only collect it but also make sense of it."

Privacy issues, and the increased adoption of cloud-based databases, will shape the future of how far and fast data fingerprinting, incorporating big data, can evolve.

"In terms of fraud, the cloud is currently more about vulnerabilities than it is about ways to prevent fraud, but that could change," says Tan.

For the future, Tan's research will see him look at innovation, and working with suppliers who are finding new ways to "make sense of big data sets".

"Mobile technology, the cloud, ERP – all these buzzwords are going to be synced with big data to produce the next platforms for innovation," he says.

"The outcome we look for is to find synergies between suppliers, customers and vendors in the ecosystem and take the fight against online fraud even further."

The study authors acknowledge the assistance of Trustev CEO Pat Phelan, Donal Cahalane, formerly CMO of Trustev, and Michael Cahalane at UNSW Business School.