Research firm Cybersecurity Ventures predicts that global cybercrime costs will reach US$6 trillion annually by 2021, up from US$3 trillion in 2015. CFOs need to arm themselves with powerful, innovative and cost-effective technology solutions to protect their businesses.
Machine Learning (ML) is a fast-developing area of computer sciences that can help them identify fraud quickly and better protect their organisations' finances.
Machine Learning to assist decision-making
In its 2016 cybercrime report, research firm Cybersecurity Ventures says enterprises targeted by online criminals can suffer damage and destruction of data, stolen money, lost productivity and theft of intellectual property, personal and financial data, embezzlement and fraud.
"Traditionally, cyber security has been focused around protection and prevention. But with the growing complexity of the threat landscape, businesses are realising the need to take a more proactive approach," explains Ned Baltagi, Managing Director, Middle East & Africa at SANS, a global provider of cyber security training and certification.
"Organisations are now tapping into the wealth of data being generated by their security infrastructures and using this to identify patterns, uncover vulnerabilities and stay one step ahead of would be attackers."
Whereas traditional data analysis to identify fraud is a lengthy process that may involve forensic accounting, new computer programming techniques like Machine Learning can considerably shorten the time it takes to spot unusual patterns.
Identifying fraudulent patterns
Unlike the standard, pre-programmed algorithms of the past, Machine Learning programmes are able to « learn » and therefore re-programme themselves. They can help identify abnormal behaviour, indicate the likelihood of fraud and intervene to reject transactions.
In the case of credit card theft for instance, if a card is suddenly used 10 times an hour rather than 10 times a month, the programme will pick it up and make intervention possible.
In the UAE, where 1 out of 10 residents is a victim of online credit card fraud as reported by a recent survey by the Dubai Department of Economic Development (DED) and Visa, this could prove an invaluable tool for CFOs to better protect the employees owning corporate credit cards as well as customers.
ML can also make more accurate predictions as its algorithms are exposed to more and newer data. "Machine Learning solutions offer the ability to analyse tremendous volumes of data, and rapidly identify patterns and categorise transactions, which is why these tools are becoming extremely attractive to organisations," explains SANS' Ned Baltagi.
"Advancements in computational power and the development of algorithms that more accurately simulate human thought processes make it possible for machines to take informed decisions grounded in a tremendous wealth of experience and consider a number of factors. The speed and efficiency are particularly important here as delays can turn away genuine customers".
Predictive anti-fraud solutions
A recent pilot of Orange Silicon Valley (OSV), the San Francisco-based division of the leading telecom operator, offers an example of what Machine Learning can do.
OSV used Deep Learning (DL), an advanced form of the technology, to sift through very large quantities of telephone records in order to recognise changing patterns of fraud. It aims was to stop the offers of cheap calling rates that use IP calling and then drop onto Orange's networks for the last mile.
"This type of call fraud uses constantly-changing mobile phone numbers to route traffic. OSV's team used Deep Learning to identify those number sets to help human agents validate and disable them, and fed back the results to the algorithms to train them," the firm said.
A key benefit of Machine Learning technology is the very fact that it can assist humans in the decision-making process. This will prove particularly useful within the backdrop of a severe cybersecurity workforce shortage, with one million cybersecurity jobs open in 2016 and 1.5 million by 2019 (Cybersecurity Ventures).
"One of the best ways to overcome shortages in staffing and funding is through automation," says SANS senior instructor Dave Shackleford. "Machine learning offers insights that could help less-skilled analysts with faster detection, automatic reuse of patterns detected and more, leading to related improvements in risk posture."
With developers now able to build Machine Learning into their cloud applications, the adoption of predictive models will make CFOs' lives a lot easier.
© Oracle 2017
