Published: Sep 11, 2025
safeguarding consumers from SMS scams using generative IT
Telco networks have become prime targets for fraudsters, resulting in significant financial losses for both operators and consumers. Traditional fraud management systems are no longer sufficient to address the complexity and scale of modern fraud. In response, NCS piloted a Generative AI-driven fraud detection system that significantly reduced fraud incidents for our client. This advanced system performs real-time analysis of network traffic, compares it against known fraud patterns, and continuously fine-tunes datasets to detect fraudulent Short Message Service (SMS) activity. Following the trial implementation, our client achieved over 90% accuracy in distinguishing scam messages from legitimate ones, enhancing fraud prevention, improving operational efficiency, and maintaining compliance with privacy regulations.
The challenge:
Traditional fraud detection systems rely on static rules and are increasingly ineffective against evolving scam tactics. In 2023, the Cyber Security Agency of Singapore reported that nearly 13% of scams were likely AI-generated. Scammers now subtly alter messages, by adjusting tone, links, and structure to bypass detection filters. These messages often appear legitimate, making them difficult to flag without advanced language analysis. Furthermore, handling sensitive SMS data requires a secure, private solution, ruling out the use of cloud-based models.
The solution:
To safeguard sensitive SMS data, a secure environment was established, and open-source Large Language Models (LLMs) were fine-tuned using both real and public datasets. The system detects scam messages and extracts key insights such as URLs, fake senders, and scam types.
Snapshot of capabilities:
- Advanced anomaly detection – identifies unusual behaviour by analysing intent-based clues from text, images, and sound.
- Real-time insights – instantly process large volumes of unstructured data to detect suspicious activities across various channels and touchpoints.
The impact:
The Generative AI-driven system delivered exceptional results, transforming the telco’s approach to fraud prevention, customer protection, and operational efficiency.
Key outcomes include:
- Achieving over 90% accuracy in classifying scam legitimate SMS messages using fine-tuned open-source LLMs.
- Leveraging historical data to identify potential vulnerabilities, classify messages, and extract valuable scam intelligence, all within a secure, on-premise deployment.