road-sign-fraud

The number of fraud professionals planning to use machine learning to tackle fraud in the next three years has doubled, finds research. 

 

New research from Callcredit Information Group found that the number of fraud professionals planning to use machine learning to tackle fraud in the next three years has doubled from 11 per cent last year (2016) to 22 per cent in 2017.  In comparison, the number planning to deploy document verification to validate customers’ identity reduced by over 75 per cent in the past year (from 26 per cent in 2016 to 6 per cent in 2017).

The new 2017 Fraud and Risk report – ‘Smarter future of fraud prevention’, which surveyed over 100 fraud professionals, also found that the intent to use device checks for identity verification fell to 17 per cent this year from 26 per cent in 2016.  At the same time, the number of fraud professionals planning to use artificial intelligence (AI) and voice recognition for the same purpose increased by 7 per cent and 15 per cent respectively further cementing the move towards machine learning.

Whilst the research shows a clear intention to get ‘smarter’ when it comes to fraud prevention, the current actual use of more advanced and complex prevention measures remains relatively low. Less than one in five (17 per cent) fraud professionals use behavioural data to provide fraud insights for their business at the moment. Furthermore, only 12 per cent use location data for the same reason.

John Cannon, Commercial Director – Fraud and ID, Callcredit Information Group, commented: “Whilst it’s great to see that organisations are planning to implement more advanced fraud prevention techniques, it’s clear that the industry will need to undergo significant change over the coming years to meet these objectives. Technologies such as voice recognition, AI and machine learning can all provide more advanced ID verification methods that are incredibly difficult for criminals to replicate. This can, in turn, help to create a smoother customer journey, therefore lending businesses a competitive edge as well as protecting their reputation.”

David Birch, author, adviser and a founder of Consult Hyperion, added: “To defend consumers against the growing threat of ID fraud, we need a better way of verifying their identity. Ultimately, organisations should move away from an approach that asks people to provide copies of their passport, towards a more secure one that doesn’t risk leaving key personal information vulnerable to being exposed. Using the latest technology, digital identity verification by banks – who are already subject to strict regulations over customers’ credentials – could be one way to better safeguard consumer’s digital identities going forwards.”

Alongside a desire to get ‘smarter’ when it comes to fraud prevention, the survey also revealed a shift to new strategic priorities, with a focus on commercial rather than tactical implications. Last year, the biggest priority for fraud professionals was creating a fraud aware culture (86 per cent) but this was only chosen by 73 per cent this year. However, creating a competitive edge went up from 56 per cent in 2016 to 71 per cent in 2017, and protecting their customer brand increased to 78 per cent from 71 per cent in the same period.

John Cannon continues: “As well as focusing on implementing these technologies within their identity verification techniques, organisations should also implement machine learning and AI into every element of their customer journey and lifecycle. Only a fully joined up approach throughout the customer lifecycle will truly ensure a seamless experience and help businesses to achieve their strategic priorities.”

Callcredit recently also undertook a year-long machine learning trial looking at a number of different scenarios ranging from identifying potentially fraudulent applications through to the accuracy of underwriting decisions in predicting a customer’s propensity to pay and bad debt across a portfolio. The project pointed to potentially significant positive benefits in productive accuracy across a range of established data models. In one modelled scenario, the level of default in a portfolio of 60,000 credit cards was reduced significantly, resulting in a 10% reduction in overall bad debt. If used with other elements of the customer lifecycle, potential machine learning generated benefits could be even greater.