Australia’s Notifiable Data Breaches scheme came into effect recently, and failure to abide by the requirements will not only put organisations at risk of millions of dollars’ worth of fines, but could result in extensive reputational damage.
Understanding the new legislation and preparing a data breach response plan is one thing, but preventing a breach in the first place is another ball game entirely. Despite Australians being the most satisfied consumers across Asia Pacific with regards to post-fraud services from banks and insurance companies, an unacceptably high one in five consumers are falling prey to fraud in Australia each year.
There’s clearly more to be done, and there’s an obvious solution: implementing ever-evolving artificial intelligence (AI) technologies.
Putting the programs to the test
The financial services sector is amongst the most enthusiastic and proactive when it comes to AI and machine learning adoption, according to recent McKinsey research. And fraud prevention is a worthy test of these innovations’ capabilities.
AI, for the first time, provides a proactive approach for tackling this rapidly evolving threat by harnessing self-learning algorithms to make sense of the growing ‘data exhaust’ generated by consumers.
Traditional fraud detection methods depend on surfacing the first few instances of fraud, or the ability of human reviewers to pick up patterns in the data and then build safeguards against such fraud. With AI, we are finally able to combine human pattern-recognition capabilities with a computer’s untiring and potentially unlimited ability to look at vast amounts of data 24/7, and at incredible speeds.
The technology’s ability to analyse vast troves of data - and use smart algorithms to detect even the subtlest aberrant patterns in transactions and other user behaviour - is obvious.
Banks and credit card companies now have the ability to spot dubious activity almost instantly, sometimes stopping the fraudster mid-transaction. There has also been a marked reduction in false positives - in other words, when a card is wrongly denied on suspicion of improper use.
The benefits are clear: better and speedier fraud protection stems the growing financial losses from card fraud, which cost Australian consumers a record AU$534 million in 2016 alone.
In many countries, such losses have also triggered a new resistance among banks to refund defrauded clients - a policy that, if more widespread, could significantly shake consumer confidence.
Playing by the rules
Developers of next-generation algorithms need to keep an eye on potential regulatory issues. Machine learning produces “black box” algorithms - complex models whose inner workings are at least partially opaque. For regulators and institutions at large, legal issues might ensue if the reasons for suspecting a fraud cannot be fully explained.
Thus, fraud-detection systems need to be complemented by reporting systems that explain how they detect transgressions. They also need to be complemented with appropriate human interventions, particularly when it comes to setting the parameters within which these smart algorithms operate.
Solutions like these, however, take sustained efforts to build because they require more thoughtful design and a system-wide intent – all of which suggests that human beings may never be completely supplanted (at least not in the immediate future).
If you can’t beat them, join them
Fraudsters have long been locked in an ongoing conflict with those tasked with stopping them, making the invention of new detection techniques a necessity, merely to keep up with the global fraud syndicates’ increasing sophistication. Similarly, the notifiable data breaches amendment is an important signal to fraudsters that Australian citizens’ data privacy is being taken more seriously.
However, it is only by fusing the human capacity to design and review such an extensive system, with AI’s ability to execute this under-appropriate supervision, that we can combat the systemic threat of fraud.
It also offers an illuminating case study on both the rewards and limitations of machine learning that is proving instructive for other industries as AI percolates through the economy, suggesting that a combination of AI and human ingenuity may be the best approach to complex real-world problems.