By: Raul Azevedo

Fraud is a pervasive problem for Communication Service Providers (CSPs). When CSPs worldwide collectively lose $29 billion due to telecom fraud, they cannot afford to ignore potential solutions. AI and analytics hold tremendous potential: Gartner recently estimated that new implementations of Machine learning (ML) in fraud management have the potential to save CSPs millions of dollars, reducing fraud losses by as much as 10 percent by 2022.

So why are AI and analytics so critical in the fight against next-generation fraud?

Fraudsters are already experts
in AI and automation

There is no doubt that fraudsters are leveraging automation and technology to commit crimes faster and more effectively today than ever before. Fraud scams such as Wangiri use an auto dialer to call multiple people simultaneously and immediately hang up before they answer. The aim of the scam is to entice those who see a missed call on their phone to call back. When they do, users are charged a premium fee for the call. This fee is then syphoned off to the fraudsters. Fraudsters are also capitalizing on the rise of bot technology to conduct the next generation of subscription fraud. Fraudsters can quickly launch a bot attack that randomly tests login credentials and common user passwords to look for ways to gain control of accounts, which are then used to order devices and costly services.

Fraudsters also target CSPs directly by using machine learning techniques to detect whether a carrier has a fraud management system in place—and then target those who don’t. In addition, they will sometimes create distractions or ‘honeypots’ by committing small, easily detectable types of fraud; and then, while the carrier is focused on stopping these efforts, go after more lucrative areas.

When fraudsters infiltrate your network, the challenge is to pinpoint them in a crowd, especially when they seek to deceive your controls by imitating the behavior of ordinary customers. With so much data and complexity in the network happening so fast, however, it becomes much more challenging for fraud managers to quickly zero in on the relevant information and focus on the real fraudsters.

This is where machine learning and artificial intelligence (AI) capabilities are required. Machine learning and automated contextual analysis expedite how to respond to suspicious behavior and provide helpful background information by combining data with context and repeatedly adding the latest new data to the accumulated history. Ultimately, these capabilities help fraud managers determine the right decisions exactly when they need it.

Rise of 5G will create new vulnerabilities for fraudsters to exploit

In the growing 5G and IoT ecosystem of partners, platforms and services, CSPs will experience increased complexity that raises many fraud management challenges that need to be addressed. IoT devices have become a welcome mat for hackers. Smart homes give fraudsters new opportunities to hack: a home security system, HVAC or smart speakers are a few easy ways for the fraudsters to obtain a customer’s account details and make additional fraudulent charges. Consider this example: a smart washing machine should connect to the local grocery store to order detergent when supply is running low. Instead, if hacked, this same washing machine could now be making calls to premium phone numbers, with the charges being directed back to the customer. Yet it’s not only the devices that CSPs need to guard against, it’s also the system. The IoT’s growing ecosystem of partners, platforms and services introduce added complexity that will increase the opportunity for bad actors to find new ways of breaking the system. For example, as IoT devices are equipped with eSIMs, they will welcome new opportunities for traditional types of telecom fraud to make a resurgence, such as subscription fraud, international revenue share fraud and traffic pumping.