A Provenance-Driven Approach for Detecting Revenue Leakage in Telecom

Revenue leakage poses an immense challenge in the telecommunications sector, especially given the increasing sophistication of network breaches that can lead to sub-stantial financial losses for companies. The current state of revenue leakage detection relies heavily on rule-based techniques, often necessitating extensive manual auditing of millions of records and financial transactions on a daily basis. This process is prone to potential errors and complications particularly due to the fluctuating nature of promotions and limited-time customer plans. To address these issues, this paper introduces a novel approach to revenue leakage detection. It combines rule-based methods with data-provenance strategies, integrating debugging and drill-down techniques. A dedicated provenance model is introduced, capturing critical auditing workflows enriched with contextual and semantic data. This facilitates the identification of latent revenue leakage issues and the generation of relevant alerts. Additionally, a query-based system is implemented, designed specifically to tackle revenue leakage scenarios. Evaluation against established gold-standard datasets demonstrates the effectiveness of our approach. Not only does it accelerate automated leakage detection, but it also achieves remarkable accuracy in the cases assessed.