It appears service providers are beginning to recognize the benefits of machine learning.  WeDo Technologies’ conducted its own telco fraud survey in the fall of 2016. The survey collected responses from 14 North American communications service providers.  Twenty-one percent stated that they are currently using some form of machine learning or artificial intelligence (AI) for fraud management, while nearly twenty-nine percent revealed that they have plans to do so in the future.  Machine learning techniques are especially suited to stopping fraud, because they can be used to identify unusual patterns and correlations from disparate data sources.  By combining modern distributed system architectures with the ability to ingest massive amounts of data from new data sources, service providers are opening a new window into fraud detection that never existed before.

Digital risk profiling - the latest weapon in the fight against fraud

Telecom fraud is increasing in complexity, and the types of information we gather – our fuel for fighting fraud - needs to change as well.  With valuable external data sources at our fingertips, it would be foolish to ignore the benefits this information can provide.  The use of both external and internal data sources allow service providers to gain a deeper understanding of suspicious activities, identify patterns, and detect unusual transactions. Service providers can now leverage Big Data, the latest data mining techniques, and the power of machine learning, to create instant digital risk profiles that can help combat the growing threat of different fraud types, especially subscription fraud.  Subscription fraud is when a fraudster signs up for services with no intent to pay. It constitutes about 40% of all telecom fraud today, and it is seen as a "gateway" type of abuse because it is often the initial way fraudsters begin their broader attack on a network operator. Therefore, stopping subscription fraud has a carry-over effect of reducing other types of fraud as well. 

By tapping into the hidden value of social media and other freely available information gathered from the web, data can be utilized to create or enhance a person’s risk profile, establishing an effective way to flag high-risk individuals or businesses before they even become customers. There is a growing amount of publicly available digital data spread across various sources throughout the web on practically every individual. CSPs can leverage this information, but attempting to find and collate it all manually is inefficient and costly.

With the right fraud management tools and the latest advances in machine learning, pertinent information can be gathered and analyzed in near real-time to create a risk profile of a particular user or company in question. The main goal is to provide a statistically sound, probability-based view of the predicted risks they pose.

A Digital Risk Profile can be created by analyzing publicly available information, which can include social media activity, education, demographics, and more. Digital risk profiles can help flag suspected fraudsters by enabling fast, evidence-based decisions that dramatically increase the speed and efficiency of threat research and analysis. It allows CSPs to find hidden connections within new and emerging threats, and improve identity verification by spotting current and new customers who are actively manipulating their identities in the marketplace. Another benefit is that, with today’s technology, it can all be done over the cloud, with practically zero implementation time.

Telecom fraud isn’t what it used to be. Managing today's threats require the latest tools and the right partners. There is too much at stake to go it alone.