Using ML for Mobile Roaming Fraud Prevention

Using ML for Mobile Roaming Fraud Prevention

A leading telecom provider with millions of customers on the network was enjoying steady growth in their network usage but there was a concern and a major one. The Telco revenues weren’t reflecting the growth despite the increasing network traffic. The technology and business teams checked various possibilities that signalled probable fraudulent activity within the network.

Sincera Technologies was shortlisted and invited to help examine the suspicious activities and recommend the right solutions to address the challenge. The Sincera Cloud solutions team and 1Data, Data Analytics experts collaborated on this project.

The team proposed and implemented a real-time fraud analysis framework, significant process checks, and migration of services to the cloud. This helped the Telco realize significant reduction in fraud by identifying anomalies and outliers and achieved significant cost savings.

The Telco had billing systems in place to monitor & analyse network traffic, recognise patterns, detect suspicious activities and identify potential fraud. These BSS systems were old and outdated when it came to detecting modern fraud and revenue leakages. The Telco was in the process of modernizing the existing data pipelines but needed something extra to identify patterns, detect and flag fraudulent activities and stop such customers from using the network who were not paying for the services.

Sincera experts stepped in and proposed a vital process overhaul combined with machine learning-supported data analytics for the customer’s BSS systems and processes.

Sincera introduced real-time analytics to detect and reduce fraud by an exponential margin which would help plug the revenue leakage and at the same time proposed the use of serverless and automated cloud solutions to reduce costs as well. This strategy helped the telco improve revenues and save costs at the same time.

Sincera established, the primary goal is to develop a robust fraud detection system using machine learning and data engineering on AWS.

To achieve this, Sincera partnered with AWS and got a detailed understanding of Telco’s existing systems and planned the process integration & system migration to the new modern cloud technologies.

A cross-functional strategic team of data engineers, machine learning experts, and cloud architects were brought together.  A series of collaborative planning sessions took place to have a deep understanding of the project, from data ingestion to real-time fraud detection.

Key communications were established to all the stakeholders at Telco, vendors and partners’ teams to secure buy-in to the transformation and move swiftly into the execution phase.

The team proposed to achieve the outcome by automating and streamlining the management of rule sets, removing the burdensome operational overhead, and enabling more efficient and cost-effective fraud prevention measures supported by machine learning. This solution was built using AWS’s cloud-native, event-driven architecture.

Sincera’s DataOps and DevOps teams were able to process data in real-time to generate alerts for potential fraud by using AWS-managed serverless components.

Sincera designed a real-time data ingestion pipeline using AWS native services. Network usage data was ingested directly from the SVC provider. Automated Data processing, metadata generation, and data cleansing operations were carried out. The formatted data was validated, enriched, and summarized. This enabled the creation of detailed tables and queries for further analysis.

The next step was to build and deploy machine learning algorithms for focusing on detecting anomalies and outliers in the data. AWS Sagemaker facilitated the running of Jupyter notebooks. To support this Sincera created a Jenkins-based DevOps pipeline to provision the necessary AWS infrastructure.

Sincera proposed solutions to rewrite the processes, deploy machine learning, upgrade to AWS cloud and manage real-time analytics delivered remarkable outcomes.

Sincera successfully implemented an event-driven, real-time architecture, performed resource optimization, reduced cloud spending and hidden maintenance costs by enabling resource consumption as a service for the telco.

Sincera took care of cost efficiency within the whole solution by leveraging serverless and automated cloud solutions to avoid the overhead of managing on-premises infrastructure. In the absence of realworld fraud data, our system leveraged unsupervised algorithms to make advanced and accelerated anomaly detection a reality.


The battle against fraud is a critical priority for any telecom operator, with potential revenue losses that could run into millions, making it imperative to plug and stop gaps. This case study highlights our comprehensive approach to fighting fraud, leveraging cutting-edge cloud technology and strategic processes & methodologies to achieve significant results for the Telco.