Fraud is, unfortunately, still rife throughout the telecommunications industry, and has previously proved remarkably resistant to efforts to stamp it out.
Indeed, the TM Forum says fraud remains one of the most significant threats to operator margins and security today, noting that fraud attacks “occur continuously on even the most monitored networks, forcing operators to constantly improve their methods of protection”.
We are all, by now, fairly familiar with the types of fraud and scams that have been perpetrated by criminals for years. There are the missed call scams on mobile or fixed lines that persuade you to phone a premium rate service. Or the scams that offer a “free” ring tone and then sign you up to a subscription without your knowledge.
Now that mobiles phones and smart devices often hold personal information such as access to emails and mobile banking apps, the situation has become even scarier.
As well as the cost to users, fraud is a major financial burden for operators. For example, the Global Telecoms Risk Management Global Survey carried out by Neural Technologies for 2016 estimated that operators faced an estimated global loss of $294 billion resulting from uncollected revenue and fraud. Operators responding to the Communications Fraud Control Association’s 2017 survey also reportedly estimated that they lost $29 billion to fraud.
The question is, will artificial intelligence (AI) not only help to mitigate fraud, but also ultimately ensure that fraudulent activity — if not completely eliminated — is at least reduced to a negligible, and manageable volume? For sure, fraud detection and prevention and revenue assurance are now regularly cited as primary use cases for AI and machine learning — and there is a good reason for this.
For one thing, telecoms operators face unprecedented demand on their networks and are no longer able to keep up with the resulting proliferation of fraud and other types of unwanted behaviour on their networks. The traditional systems based on rules and established thresholds are no longer fit for purpose. That’s because telecoms fraud now evolves at such a rapid rate it is becoming impossible to keep up by relying on the old methods. This situation will only get worse as billions more IoT devices are connected, ushering in huge volumes of unsecured data onto operator networks.
Machine learning has already proved effective at fighting fraud because it is applied on a massive scale to determine the characteristics of traffic and identify anomalies that could end up being fraud. The technology is used to develop and train algorithms to monitor for such anomalies.
Many now believe that broader AI techniques linked to a deeper analysis of data are must-have technologies for operators to enable them to detect fraud faster, and also identify future risks that are likely to be even more complex. For example, AI can enable powerful automation and operational efficiencies at a massive scale, and is able to handle and streamline the management of millions of customer or network data points.
One thing is for sure, telecoms operators have no choice but to introduce new methods to mitigate rising levels of more complex threats if they are to protect their networks and business in future. As fraud schemes constantly evolve, a more adaptive approach for the identification of fraudulent activity is now necessary, and AI provides the tools to create such an approach.