Can AI save us from losing a war against scammers?

Gamifying to win is a trait that is deeply embedded within all of us. A young child knows exactly how to socially manipulate their parents to get that extra scoop of ice-cream, or more time on their devices. Thankfully, most learn, and apply, boundaries of right and wrong in their day-to-day engagements while concurrently trying to learn the subtle tells of when they are themselves subject to another’s nudge. The uncomfortable reality is that we are all subject to being played, scammed and manipulated — from the innocuous gifting to a child more than they should have, paying more than is necessary to being cheated of one’s life savings.

The relentless technological advancement happening around us is not limited to ‘the good guys’. The world of scamming has tremendously evolved with the progress and adoption of readily accessible technology. From the so-called Nigerian scams of “I am the heir to a diamond mine and need your help only for 250 dollars” to highly sophisticated and personally crafted messages that are borderline indistinguishable from their authentic equivalent. These are designed to manipulate our most basic instincts to drive towards what is then retrospectively seen as an irrational action. It is important for us to take note that unlike ‘us’, scammers are not having roundtable discussions on governance and compliance in AI or technology at large. Whether innovation should be balanced with control.

Today, the measures for scam prevention fall largely into two areas. First, the customer end, which is actioned through education, warnings-alerts, device-level limitations and various other friction points. Customer education is a critically important facet and must have dedicated resources and effort. The first truth we need to embrace is that no matter how informed, educated or careful a person is they will remain subject to being manipulated. Scammers are artists and we are their canvas.

Second, is broadly a transactional layer. It encapsulates a crucial aspect of the fight against scams. This layer encompasses various entities, including telecommunication companies (telcos), financial institutions, regulatory bodies, and other relevant parties. The primary focus of the transactional layer is to leverage information about known scams and apply it to transactions in real-time. This analysis aims to identify and flag potentially fraudulent activities, providing an additional layer of protection for consumers and businesses.

At the core of the transactional layer are topologies, manually crafted rules which try to capture the characteristic properties and patterns associated with the respective scams. These topologies are derived from reviewing historical incidents and intelligence gathered from various sources, such as law enforcement agencies, fraud detection systems, and consumer reports. By analyzing transaction data against these topologies, systems aim to identify suspicious patterns and anomalies that may indicate a high likelihood of a scam.

Topology, rule-based, approach are innumerable rules of past irregularities or risk indicators that are built on top of each other almost like a house of cards. The numerous interconnected rules can quickly become ineffective due to evolving bad-actor behaviors and system knowledge. With 90 percent or more false-positives, we are to the glee of the bad-actors targeting our focus on noise.

Furthermore, there is a growing trend towards collaboration and information sharing among organizations involved in the fight against scams. Through initiatives such as industry-wide fraud prevention networks and public-private partnerships, organizations can pool their knowledge and resources to create a more comprehensive and effective defense against scams.

The belief that we possess a clear understanding of the characteristics of scammers and scams is a fallacy. If we had this knowledge in advance, we would witness a decline rather than an increase in scams. The reality, however, is that we often only become aware of these properties after a scam has already taken place and victims have come forward. This leads to a fundamental appreciation that it is not a matter of “if” a scam will occur, but rather “when”. The reason for this situation lies in the fact that the industry tends to treat the symptoms rather than addressing the root causes. The scammers would have already successfully compromised the financial infrastructure before attempting to engage a victim. After all why try and scam money if the illicit funds cannot be received.

How can Artificial Intelligence (AI) save us from what feels like a losing war? AI is the ability to see things differently — a family of methodologies to find and generate patterns, including the irregularity of patterns. It holds the ability not only to detect these patterns but to potentially predict them in advance. The true win for AI is the ability to be behavior focused — anomaly detection coupled with the ability to look for the subtle changes in behavior that may indicate risk or distress.

We must expand our collective appetite beyond treating painful symptoms. We must embrace AI driven capabilities to proactively and continuously identify when the financial infrastructure has been compromised by bad actors. Current strategies for scam prevention focus on bubble-wrapping the end users. It fails to prevent the scam and does not go to the root of the issue. Using AI, we have the ability to elevate defense and protection mechanisms, not only for the consumers, for the financial institutions, but also for the scammers. We can deploy ‘Minority Report’ like capabilities to identify and mitigate scammers before they scam. The value of true application of AI is that no topology is needed in advance. Anyone remember Sesame Street’s “One of These Things (Is Not Like the Others)”?

Imagine a world where scammers still try to manipulate victims but now even if they succeed the scammers are prevented from accessing and moving the funds because proactive predictive measures have identified the true intent of the receiving accounts or store-of-value. A few organizations have embraced such solutions in full scale operations with demonstrated results of 90 percent true-positive. This means nine out of 10 accounts preemptively identified are validated as bad-actors, or compromised accounts, with the intention to scam in the future. Contrast this to today’s widely adopted traditional transaction-monitoring topology driven approach which hovers around the 90 percent false-positive in catching an illicit transaction. This means nine out of 10 transactions that are presumed to be potentially illicit, are in fact legitimate.

AI enables us with the tools and capabilities, the responsibility lies with us to have the imagination and appetite to go beyond current practices in achieving new possibilities. Can AI save us from the losing war with scammers? Perhaps, if we start using it.

Dr. David R. Hardoon is the Chief Executive Officer of Aboitiz Data Innovation and Senior Advisor for Data & AI to Union Bank of the Philippines. David is the first appointed Chief Data Officer at the Monetary Authority of Singapore.

The opinion was generated by a human. During the course of writing this opinion, several scam calls and emails had been received.