Stopping financial crimes with machine learning and AI
Andrew Simpson, Chief Operating Officer at CaseWare Analytics
When people talk about financial crimes, most think about activities that can impact them directly such as credit card fraud, tax evasion and elder abuse. But in reality, financial crimes include much broader offences - such as human trafficking, drug trafficking, money laundering and terrorist financing that can have lasting effects on society.
One of the problems when addressing financial crimes is that organisations are taking a retrospective view. They are chasing resolutions for these crimes rather than preventing them from occurring in the first place. But things are changing and technologies such as machine learning and AI are helping to propel these changes.
Today’s approach to detecting financial crimes
“The big problem is that companies think they need to establish rules, policies and procedures to prevent fraud,” says Andrew Simpson, Chief Operating Officer at CaseWare Analytics. “But today’s criminals are much more sophisticated and are able to circumvent these business rules.”
According to Simpson, businesses need to take a more dynamic approach that includes both business rules as well as machine learning and AI in order to learn from evolving criminal behaviour and deliver a more sophisticated and effective approach to dealing with financial crimes.
This approach to fraud and crime detection is not limited to just large multinationals or financial institutions. Fraud plagues small organisations just as much. According to the Association of Certified Fraud Examiners (ACFE), 30 per cent of fraud cases occurred in small businesses and 60 per cent of small-business fraud victims didn’t recover any of their losses.
The same report also found that small businesses have fewer anti-fraud controls in place than large organisations, making them especially vulnerable. “Machine learning and AI present an amazing and cost-effective opportunity for anti-fraud departments in small companies to evolve their programmes and increase the effectiveness of their efforts,” says Simpson.
Case study: Lottery scams and elder abuse
Machine learning is also an effective approach to deal with crimes such as lottery scams and elder abuse. “Imagine your 75-year-old grandmother buys a lottery ticket every week and suddenly she receives a call that she’s won £10 million,” says Simpson. “But to claim her prize she must pay a £1,000 processing fee. To an elderly person this may seem reasonable, so she will wire the money.”
But, as it happens, in many cases, the perpetrators don’t stop after the first wire transfer. “Instead of being awarded her prize, she might then receive another call from the scammers telling her that there is a £5,000 tax fee that she must cover in order to receive her £10 million prize. Before you know it, your grandmother has been swindled out of thousands of pounds.”
With a rules-based approach, the initial £1,000 wire transfer may have gone undetected if the rule was to only flag any transfers of more than £3,000. However, with AI and machine learning software, the system will have been trained from the customer’s historical data that the wire transfer was suspicious, enabling the financial institution to pause the transaction and investigate further before clearing it.
In the event that the initial transaction was only flagged as fraudulent after it was cleared, with machine learning applications could then learn that transactions to that particular recipient or IP address were illicit and to halt future transfers.
Looking ahead over the next five years, Simpson believes that the fight against financial crime will shift towards more real-time detection and prevention, rather than a reactive investigative approach. “You will see more AI and cognitive analytics being deployed, not only for preventing suspicious activities but also investigating and making decisions about incidents,” he concludes.