BPM Analytics

Digital twin – The saviour of the telecom fraud management landscape

Telecommunications is one of the most dynamic industries, helping us to stay connected by leveraging a nexus of the latest technologies. Telecommunication goes hand in hand with technological evolution, all the while adhering to high technical standards, says Anand Chandrashaker, senior domain principal, digital transformation services and Apurva Prashant Patil, senior analyst, digital transformation services (fraud risk assurance compliance) at Infosys BPM.

However, the industry’s vulnerability to fraud poses a serious and ubiquitous problem, resulting in huge revenue loss. According to a study by Forbes, in 2021, the global telecom industry suffered a loss of 2.2% with a monetary loss of [$40 billion (€36.57 billion)].

A new approach to data processing

Fraud management essentially follows two stages – fraud detection and fraud prediction. Currently, all solutions addressing fraud management follow a one-size-fits-all approach and work on known fraud stimuli for all customer bases. Telcom companies are mainly dependent on a rule-based fraud detection approach, which is more reactive than proactive.

The implementation of 5G and edge computing inevitably results in the generation of tonnes of data, almost every minute. Processing this incredible amount of data solely through a rule-based system is time-consuming, somewhat impractical, and can even be detrimental to customer experience. The deployment of artificial intelligence (AI) and machine learning in this sector has been known to address these issues to some extent, while the full range of their capabilities is being explored.

The rise of the digital twin

The approach introduced by these technologies focuses on the potential victim, using prediction and modelling to identify the reasons why fraudsters choose a victim. The focus of these models is on pinpointing the probability and reason why any real-life entity ends up becoming a target for fraudsters. Digital twin technology plays a major role here.

Digital twin technology has the capability of ingesting large volumes of data and running models and business rules simultaneously, thereby allowing the prediction and detection of anomalies. It adds context to the monitored activities, which enhances prediction speed and accuracy and reduces dependency on data from each fraud prediction to gain knowledge on potential victims by creating virtual replicas of systems in near-real time.

Let’s assume dual paths, out of which one path allows data stream into ML models for fraud prediction and model training, and the second path derives the context to update the model through digital twin technology using both near real-time and historical data. 5G speeds up the process in many ways like reducing latency and increasing throughput to capture near real-time data. Additionally, 5G helps in experimenting, testing, and optimising digital twins by analysing massive amounts of data.

How the digital twin is a trump card for fraud mitigation

This concept can be better understood with an example. 5G deployment helps in the effective implementation of connected cars and related applications. The dynamic nature of vehicle networks with the heterogeneity of wireless infrastructures of connected cars makes resource management and low latency communication requirements a challenge. Digital twins can be used to analyse the overall performance of connected vehicles and 5G network devices. AI will be used to predict vehicle and network performance under dynamic conditions, detect suspicious activities, identify problems at an early stage to avoid vulnerabilities and provide or suggest immediate solutions. Technically, the digital twin helps facilitate core functions such as the drafting of service level agreements (SLA), validating data rates with real call traffic, handling complex handover scenarios, monitoring radio performance, etc.

Building a digital twin – A slow but steady process

Digital twin implementation has its challenges and it is not yet deployed widely in the telecom industry. It requires an amount of data and computing resources. Integrating data that comes from various sources, and vendors in different formats can be very challenging. Accuracy and completeness are critical parametres for a digital twin. Data quality is a crucial factor and may be affected by device accuracy, data loss, and human errors. Developing a high-fidelity digital twin is complex due to the complicated nature of telecom infrastructure, the knowledge-intensiveness of the process, and the need for subject matter expertise. The implementation of digital twins comes with a high investment cost and time commitment.

The digital twin has the potential to enhance the governance for AI and usher in extreme automation in the future, by bridging the gap between processes and automation through reliability and data-driven insights. It will certainly help in 5G design and implementation using a virtual approach. Further, it opens opportunities in fraud prediction and mitigation alongside the ML techniques that enable the detection of potential threats on 5G users. There is a long way to go till we fully utilise digital twin technology in telecom fraud management, but the sector has started laying the foundation for a secure tomorrow.

This article was first published on Vanilla Plus

Recent Posts