Fraud is a big problem in the cellular networking market, and machine learning is one potential solution to the problem.
Fraudulent usage of cellular networks costs the industry an estimated $38 billion a year, according to the 2015 Global Fraud Loss Survey by the Communications Fraud Control Association (CFCA), an international organization that promotes revenue assurance, loss prevention, and fraud control in the industry.
The CFCA says fraudsters use methods including PBX hacking, subscription fraud, dealer fraud, service abuse, and account takeover to steal from service providers.
Current fraud detection approaches in the industry rely on static rules with pre-set volume or frequency thresholds, said Ole J. Mengshoel, associate research professor in the Department of Electrical and Computer Engineering and director of the Intelligent and High-Performing Systems Lab at Carnegie Mellon University.
"This means they can only detect fraud types that conform to known configurations," said Mengshoel, who has authored a research paper on the topic. "Fraud specialists are constantly working to uncover new fraud types, but modern cyber attacks evolve faster than analysts can write rules to detect them."
Adaptive artificial intelligence (AI) and machine learning can help address these weaknesses and reduce fraud in the cellular services market.
"Innovators like Facebook, Google, and LinkedIn have pioneered big data and machine learning approaches to protecting their subscribers and gaining insights," Mengshoel said. "New machine learning approaches start from the position that the only way to detect anomalies in real time is to apply machine learning at massive scale."
The combination of supervised and unsupervised machine learning makes it possible to analyze massive amounts of data and alert fraud analysts in seconds, Mengshoel said.
Products are already on the market that combine deep packet inspection of big data with supervised and unsupervised machine learning to perform network analytics for fraud, anomalous traffic, and other network behaviors in real time, Mengshoel said.
"Their real test will be which vendors are able to perform network analysis on the data plane as well as the voice network," he said. "More and more traffic, and therefore more and more fraud, happens on the data plane."
The research paper by Carnegie Mellon and Argyle Data, a provider of big data/machine learning analytics technologies for mobile providers, describes how real-time anomaly detection can be used for near-instant identification of fraud.
The report shows how current solutions cannot address issues on the data plane, and why in the future gaining visibility into the characteristics of data usage will be paramount. Because of the vast amount of data flowing across telecoms networks, big data analytics capabilities and the ability to analyze these using advanced machine learning are essential.
In their research, Mengshoel and coauthor David Staub, data scientist at Argyle Data, validate a supervised and unsupervised machine learning-based approach that automatically learns the difference between normal and anomalous call patterns based on usage data.
A solution to fraud can't come soon enough. As the paper notes, fraudulent or unacceptable use of cellular networks is a growing threat for both network subscribers and operators, and fraud schemes are constantly evolving. In this environment, the report said, a sophisticated, adaptive approach for identification of criminal activity is needed.