How Axiata uses data analytics to thwart telecom fraud
TELECOMMUNICATION or telecom fraud is fast becoming one of the biggest sources of revenue loss for telco providers everywhere.
Being a low-risk alternative to traditional financial crime, the rise in telecom fraud costs the industry more than US$32.7 billion annually, according to one estimate.
While the fraudulent activities itself are not new, it poses new challenges for law enforcement and telco providers everywhere and many organizations in the space realize that a novel approach is needed to address the issue effectively.
One Malaysian telco conglomerate, with extensive operations within Asia, decided to pursue precisely that when it chose to leverage data analytics to thwart telecom fraud.
Axiata Analytics Centre’s AI Head Ravi Kumar Madavaram explained that charges for international calls vary, depending on the country of origin and the destination, as well the trunk used to connect the call.
“For example, a call from the Middle East to Nepal is more expensive than a call from Malaysia to Nepal. Because of that, people route the call from the Middle East, to Malaysia to call Nepal,” he said during his presentation at the recent Malaysian Institute of Accountant’s Forensic Investigation and Fraud Analytics Conference.
By doing so, they end up paying far less to local providers, which is a massive problem for Axiata’s subsidiary in Nepal, NCell, as up to 25 percent of its revenue comes from overseas Nepalese workers calling their families at home.
“We have very little information about the parties that are making these calls, but we need to identify them with the data we do have,” Madavaram said.
This is where data analytics comes in handy for the company.
Combatting telecom fraud with machine learning
Using data analytics, Axiata captures all the data available on these calls to review it against a specific set of thresholds and look out for particular signatures and flags.
“One of the key challenges in fraud is to continually think of how to identify different fraud patterns and subsequently stop it,” said Madavaram.
For example, one could identify a particular trend or pattern and block it, but the perpetrators will easily switch and try a different method, he said.
However, the challenge could be overcome with the use of machine learning (ML).
Madavaram explained that once the different traits of fraud are fed into an ML-based model, it can highlight potential activities that could be further investigated.
He added the ML algorithm implemented by his team in Axiata does not require human intervention when it comes to setting the rules and parameters.
“The model itself picks up trends and flags potential fraud,” he said.
Telecom fraud is not the only area where data analytics is being deployed within Axiata. The company has integrated data analytics as part of its internal audit process to enhance the accuracy and efficiency of its audits, among other things.
Axiata currently has a 50 people analytics team supporting all its regional brands and business units, including its e-wallet platform Boost.