Skip to content
How NT Used Machine Learning to Tackle Simbox Fraud - Neural Technologies
Neural TechnologiesFeb 14, 2022 2:20:02 PM4 min read

How NT Used Machine Learning to Tackle Simbox Fraud

Advanced Solutions in Action for Simbox Fraud and Sim Boxing

Simbox fraud, or sim boxing, is a multibillion-dollar challenge for communication service providers (CSPs), reducing call quality and customer experience and triggering significant revenue losses.

The Communications Fraud Control Association (CFCA) estimates that mobile operators globally incur USD28.3 billion in losses through interconnected bypass frauds, with simbox fraud accounting for USD4.27 billion of these losses. 

The Association is not alone in highlighting the impact of sim boxing, with the Revenue and Assurance Group (RAG) estimating that simbox fraud could cost CSPs up to an eye-watering USD6.86bil annually.

With the huge potential losses triggered by simbox fraud, Neural Technologies has been working with partners around the world to mitigate this threat with advanced, ActivML machine-learning-driven solutions for telecom operators. In our latest implementation, we’ve been working with major telecoms providers in Europe and the Middle East to significantly reduce their exposure to losses from simbox fraud.

Fraudsters are smart and committed individuals, and often attempt evolving strategies to target CSPs and their customers. They’re motivated to maintain their illicit profits through fraud, so as new simbox detection techniques are developed, they’re quick to evolve their operations to counter them. This same tactic is taken by manufacturers of simbox servers, whose profits rely on providing updated hardware that can deliver success for fraudsters. 

Neural Technologies’ ActivML technology delivers an evolving protection against simbox fraud, built on a data-driven approach which adapts to address emerging fraud threats. The self-learning technology can recognise and detect fraudsters’ own adapting strategies, and thus provides a robust and evolving protection against simbox fraud for CSPs. 

Sim box fraud explained

Sim box fraud involves fraudsters obtaining large numbers of SIM cards from mobile providers, then incorporating them into a device known as a GSM gateway. The fraudsters then route international calls using voice over internet protocols (VOIP) to the GSM gateway device which completes calls into the mobile network, tricking the network into thinking it’s an internal network call. This bypasses the higher costs that would be charged for international or national calls, and incurs significant lost call revenue for the CSP.

Simbox fraud is one of the most significant risks to CSPs, and one of the key areas of telecoms fraud coverage provided by Neural Technologies. Addressing the evolving nature of this fraud is a critical need to maintain protection as fraudsters’ tactics adapt. 

In some extreme cases it is claimed that bypass fraud can account for a 50% reduction in international termination revenues, and reductions of USD250,000 per month for CSPs is not uncommon. Some operators have reported losses as high as USD200 million per annum as a result of sim boxing. 

Neural Technologies’ telecoms customers were facing significant challenges from simbox fraud, leading to substantial lost revenue and unwelcome profit impacts. 

The Neural Technologies solution

Neural Technologies’ advanced machine learning (ML) and artificial intelligence (AI) solutions offer a powerful end-user configurable platform to combat simbox fraud, leveraging sophisticated analytics with Explainable AI and automated actions to eliminate sim boxing fraud threats.

We utilized a structured profiling ML model which was fed with a range of typical usage data, creating a rich dataset as the foundation to rapidly identify events which correlate with suspected simbox and bypass fraud. The rich dataset included just a small subset of fraud activities, providing the basis for the system to accurately identify high-risk activities while avoiding any risk of undermining service provision for genuine users.

Our sophisticated machine learning model was able to use fully automated structured profiling analysis of key data points such as number of calls made, duration of calls, number of international calls, overall usage profile, and other critical parameters to identify different types of activities.  This is then coupled with classification modeling to separate fraudulent activity from genuine users.  This hybrid ML approach also allows the system to identify evolving anomalous behavior that falls outside the norm, and explain suspicious activity for further investigation or link to new fraud methods. An example was a new type of camouflage calling activity introduced by fraudsters to avoid detection of sim boxing.

Unlike a traditional, static fraud management system, the AI/ML approach of Neural Technologies ActivML provides an adaptive and flexible solution that not only identifies a rigid list of fraud activities, but actively evolves to detect emerging fraud methods and types in real-time.

AI/ML solutions ensure customers’ fraud management systems are not only empowered to identify frauds following traditional methods or patterns, but also able to quickly identify emerging anomalous behavior related to new fraud risks. Automated responses can be established to eliminate the threat before revenue losses occur, while high-risk or suspicious cases can be quickly flagged and escalated to analysts for a full review supported by Explainable AI.

Neural Technologies’ solution to simbox fraud provided a powerful opportunity for our telecoms customers within our Optimus RP fraud management system that delivers on key principles of:

  • Wide coverage. Flexibly incorporates data across an unlimited range of formats for a diverse set of fraud scenarios.
  • Rapid detection. Sophisticated configurable analysis offers rapid responsive identification of threats while eliminating false positives.
  • Adaptive response. Dynamic behavioral profiling and AI/ML techniques evolve to identify new and emerging fraud types.
  • Clear oversight. Calculates and delivers clear transparency on estimated losses, rate of loss, and savings, for demonstrable ROI.
  • User-friendly data. Provides Explainable AI data visualization and reports for proactive and informed fraud management.

Simbox fraud remains a critical challenge for CSPs, but our work with major telecoms operators reveals how advanced machine learning and artificial intelligence solutions powered by Neural Technologies’ ActivML technology and extensive industry experience can help eliminate that threat even as it evolves in attempts to avoid detection.

Find out more about our ActivML-powered Optimus RP Fraud Management solution or contact Neural Technologies today to start your company’s journey to better revenue protection.

RELATED ARTICLES