Every year, people and organizations are losing billions of dollars to fraud. So it’s no surprise that organizations are also investing billions in technology that can help prevent this fraud and save them and their customers’ money.
Currently, artificial intelligence is the leading technology used to defend against fraud. AI is being used to analyze customer data and report potential fraud that it identifies. Further, AI is combating scammers by detecting scam emails, phone calls, and websites, warning users that they should proceed with caution.
In Australia alone, over 100 million dollars was lost in 2018 to fraud. And it is now said that the global fraud loss is in excess of 4 trillion dollars each year. So it’s obvious that a lot of work needs to be done to stop this number rising. Artificial Intelligence is currently the leading prevention method and banks and other organizations have seen large decreases in fraudulent activity since implementing it.
Unfortunately, it still isn’t perfect. But with enough time and research, there is hope that the global fraud loss will be much lower.
Artificial intelligence has the power to process data exponentially more quickly than humans are able to. This advantage allows AI to analyze users typical transactions, shopping locations and other useful data to detect when a fraudulent charge may be taking place. Equally important is the possibility to detect when fraudulent charges may occur in the future by identifying new methods of fraud.
Organizations attempting to stop fraud will usually start with a set list of rules that are hard-coded into their system. These rules might be something like “If a credit card has multiple large charges within a short period, flag the account”.
Unfortunately, these rules alone are not enough to stop fraud. This is where AI is able to step in and do a better job.
How Does This AI Work?
There are two main methods for teaching a fraud detection AI what to do. That is, supervised machine learning and unsupervised machine learning.
As the name suggests, supervised machine learning algorithms are basically told by the engineers what kind of decisions they should be making. When machine learning algorithms are subject to supervised learning, they are given previously collected data that has been manually labelled fraudulent or not. The data will help train the algorithm to detect fraudulent activity that is similar to that which had been previously detected.
This type of AI is excellent at preventing older and currently circulating methods. However, it is unable to find new fraudulent methods that criminals are coming up with. This is where unsupervised machine learning shines.
Unsupervised machine learning algorithms are much more complex to create. Essentially, developers are aiming to have the algorithm make decisions about potential fraud on its own, without being specifically told what to look for. These algorithms will process large amounts of unlabeled customer data and will learn to recognize new fraud attacks and stop them in their tracks. Put simply, the algorithm does this by analyzing patterns in the data and reporting any anomalies it finds.
Using just one of these models alone is often not enough, however. This is why organizations are now focusing on finding a good mix of the two types of AI. When PayPal started using machine learning algorithms, they reported a fraud rate of 0.32 percent, while others in the industry averaged 1 percent higher, at 1.32 percent. So it’s clear that these methods are effective. It’s just a matter of implementing them in more places and continuing to train and perfect the algorithms.
How Else is AI Helping?
Not only is AI being used to stop fraud as it happens, but it is helping prevent it from occurring at all.
You may be familiar with Captcha technology. This is a little box prompting you to type in what you see, or perhaps click on certain images. It may be annoying, but this is drastically helping to prevent scams by stopping bots from accessing websites. Additionally, it can help stop spam from being posted in public forums and websites making it harder for criminals to get their scams out to the public.
While some of these Captcha programs are simple and do not use AI, a lot of the newer models do. Google’s ReCaptcha uses machine learning algorithms to analyze data such as your location, IP address and search activity. Based on this information, it may then prompt you to click on a certain series of images that are provided to you. If your IP address has been associated with spammy activity in the past, you will often have to answer multiple questions. ReCapture makes it nearly impossible for non-humans to bypass.
Another excellent use of AI allows newer phones running Android to detect incoming phone calls that could be scams. You will be shown a warning saying something like “Suspected Scam”.
Gmail also uses both user reports and AI to detect potential scam emails and provide a warning to the user advising that the content may be unsafe.
Emails or phone calls you see with this type of message should be avoided completely.
Many efforts have also been made to prevent scam websites from being accessible. Google Chrome users may come across sites that display a warning message such as “Deceptive site ahead”. These sites have been flagged by Google because they are often attempting to steal personal information such as passwords or credit card numbers.
Another way AI is being used to stop scammers and fraudsters is by simply wasting their time.
There are several organizations that offer these services and will do everything they can to prevent scammers from talking to real people. One such service is re:scam. They use AI to reply to scam emails and waste as much of their time as possible. So far, they claim to have wasted over 5 years of scammers lives and have sent over 1 million emails. Doing this may seem silly, but the more time scammers spend talking to this AI, the less time they can spend scamming real people.
Projects like this are becoming increasingly popular and really do help. You can read more about re:scam here: https://www.rescam.org/
Written by Cameron Christensen
Cameron Christensen owns and operates the website Techstry.net. Cameron publishes the latest interesting technology news, as well as in-depth tech-focused articles.