Responding to COVID-19: insight, support and guidance
While fraud is an age old behaviour, with the rise of the digital age, criminals are finding new and inventive ways to deceive both unsuspecting victims and our public services. The Public Accounts Committee (PAC) in their paper on ‘The Growing Threat of Online Fraud’ found online fraud is now the most prevalent crime in England and Wales, with the Office for National Statistics (ONS) showing that around one in six of all estimated crimes in England in the year ending March 2018 to be fraud committed online. The cost of this crime is estimated to be £10 billion, with around 2 million cyber-related fraud incidents last year, however the true extent of the problem remains unknown. Only around 20% of fraud is actually reported to police, and in the case of individuals, the emotional impact of the crime means many victims are reluctant to come forward.
In the public sector significant efforts are being made to tackle fraud across government, however with new digital elements now emerging, progress is difficult. An example of this can be seen in the Parliamentary inquiry on ‘Policing for the future’ which found that as little as 3% of cases reported to Action Fraud may result in charges or summons. In the case of fraud against individuals it is unlikely that more than one in 200 victims ever sees their perpetrator convicted. Similarly, when it comes to local government, while the latest CIPFA Fraud and Corruption Tracker (CFaCT) report revealed an increase in the number of frauds detected or prevented – up to 80,000 – a total of £302m, in truth this represents only a fraction of the fraud committed, as estimates put the cost to the taxpayer in the billions. A movement towards hosting most transactional public services online will introduce additional challenges for both central and local government when it comes to tackling this issue.
However, just as the means of how fraud is committed has evolved, so too has the technology available to detect and catch fraudsters, and some government departments are embracing it. One great example can be seen in the UK with the Department of Work and Pensions (DWP), which has begun to use artificial intelligence (AI) to drive a crackdown on benefit fraud. This is an area of significant loss for the public purse, with nearly £3.6 billion made in benefit overpayments in 2016 to 2017 according to the ONS. By developing and launching state-of-the-art algorithms, DWP aims to detect a number of identity cloning techniques which have been commonly used by organised criminal gangs committing mass-scale benefit fraud. The algorithms look for patterns in claims such as the same phone number, or applications written in the same style. Once a claim is flagged as suspicious, a human investigator takes over to determine if the claim is in fact fraudulent.
Fraud is an area where AI can prove especially useful, as it can identify possible issues far quicker than human investigators. While this of course does not remove the need for human oversight, it can greatly increase the efficiency of investigations, and free up resources to tackle the major cases. It also enhances the toolset available to investigators by providing another means of detecting fraud, other than tips from the public. Freedom of information requests by the media to DWP have shown that over the financial years 2015/16 and 2016/17, of the 332,850 cases closed following reports by members of the public, 287,950 were found to have no or little evidence to substantiate the claim – or 87%. It will be interesting to see DWPs use of AI to detect fraud evolve, and I am looking forward to seeing the results when they next release information.
With a similar approach being taken in the private sector, it seems AI technology has been flagged as the tool of the future in the fight against fraud. Many insurance agencies and banks have been investing in AI and machine learning to take on fraud with Lloyds Banking Group, for example, using machine learning for online fraud prevention with models that can detect whether the person logged in to our online banking is a customer, a fraudster, or even a ‘bot’. Similar techniques are also being use by insurers to detect benefit fraud to flag suspicious claims. Undoubtedly, with large data sets, government is a place where AI can make a significant impact, and I expect it will likely become an integral part of how the government combats fraud. In the future, it appears this battle will no longer be fought by humans alone.
This article first appeared in Public Finance Magazine.