Research Article
Subscription Fraud Prevention in Telecommunication Using Multimodal Biometric System
@INPROCEEDINGS{10.1007/978-3-031-34896-9_28, author={Freddie Mathews Kau and Okuthe P. Kogeda}, title={Subscription Fraud Prevention in Telecommunication Using Multimodal Biometric System}, proceedings={Towards new e-Infrastructure and e-Services for Developing Countries. 14th EAI International Conference, AFRICOMM 2022, Zanzibar, Tanzania, December 5-7, 2022, Proceedings}, proceedings_a={AFRICOMM}, year={2023}, month={6}, keywords={Subscription Fraud Telecommunication Fingerprint Biometrics Face Biometrics PCA Algorithm Multimodal Biometrics System}, doi={10.1007/978-3-031-34896-9_28} }- Freddie Mathews Kau
Okuthe P. Kogeda
Year: 2023
Subscription Fraud Prevention in Telecommunication Using Multimodal Biometric System
AFRICOMM
Springer
DOI: 10.1007/978-3-031-34896-9_28
Abstract
South African telecommunications market has reached a saturation point; as a result, telecommunication companies spend most of their budget on customer acquisition and retention, and very little is spent on fraud prevention or detection systems. This spending pattern has caused an increase in fraud, making it the most significant revenue leakage in telecommunications, where the leading fraud type is subscription fraud. Subscription fraud has a direct negative impact on the company’s revenue, bonuses of employees, and customers’ credit status. Although the current fraud systems can detect subscription fraud, they cannot identify the fraudster. This enables the fraudster to commit fraud using the same or multiple identity documents during the contract application process without being detected. In trying to change the spending pattern and prevent subscription fraud, we sought to determine the impact of subscription fraud in mobile telecommunication companies. We designed, developed, and implemented a Multimodal Biometrics System (MBS) using Python, SQLite3, and JavaScript to enable telecommunication companies to capture and store customer faces and fingerprints to use them for verification before approving the contract. We used Principal Component Analysis (PCA) algorithm to reduce the dimension of the face and fingerprint images. PCA outperformed Independent Component Analysis and Linear Discriminate Analysis algorithms. To do image matching, we used the PCA-based representation for local features (PCA-SIFT) algorithm, which outperformed Scale Invariant Feature Transform (SIFT) and Oriented FAST and Rotated BRIEF (OBR) algorithms. MBS results gave us biometric matching accuracy of 94.84%. MBS is easy to implement and cost-effective. The system can help identify the fraudster, prevent subscription fraud and reduce revenue leakage.
