An Artificial Neural Network-Based Caller Authentication and Identification Algorithm in Cellular Communication Networks

• Published in: International Journal of Applied Methods in Electronics and Computers
• Main Authors: Sombo, Bem, Apeh, Simon Tooswem, Edeoghon, Isi Arthur
• Format: Journal Article
• Language: English
• Published: 31.12.2024
• ISSN: 3023-4409; 3023-4409
• DOI: 10.58190/ijamec.2024.111

Cellular communication companies have created tens of thousands of jobs within the industry. However, while business opportunities have expanded for legitimate enterprises, criminal gangs have also exploited the new business environment to further their illegal activities. Weaknesses in the authentication algorithm allow criminals to commit fraud while remaining anonymous, thereby facilitating wireless crime. This paper presents the development of an artificial neural network-based caller authentication and identification algorithm for cellular communication networks that addresses these weaknesses. Voice data were collected from volunteer participants and the Mozilla Common Voice (MCV) database. Voice feature vectors were extracted from the voice data using the Mel Frequency Cepstral Coefficient (MFCC) technique. The extracted voice feature vectors were used to train a Multilayer Perceptron (MLP) neural network for voiceprint generation. The MLP neural network architecture was optimized through Neural Architecture Search (NAS) using the Neural Network Intelligence (NNI) toolkit. The optimized MLP neural network architecture was trained with the extracted voice feature vectors to generate a voiceprint generation model. The developed voiceprint model was then deployed to create an artificial neural network-based caller authentication and identification algorithm in cellular communication networks. The developed algorithm was evaluated in the cellular communication emulation setup using accuracy, false acceptance rate (FAR), and false rejection rate (FRR) as metrics. The results of the performance evaluation of the developed algorithm for authenticating first-time users of a new SIM and identifying third-party SIM users both showed 100% accuracy, 0% FAR, and 0% FRR. These results indicate that the developed algorithm has strong potential to prevent the use of black market SIMs and deter perpetrators from using third-party SIMs to access network services while remaining anonymous. Cellular communication companies have created tens of thousands of jobs within the industry. However, while business opportunities have expanded for legitimate enterprises, criminal gangs have also exploited the new business environment to further their illegal activities. Weaknesses in the authentication algorithm allow criminals to commit fraud while remaining anonymous, thereby facilitating wireless crime. This paper presents the development of an artificial neural network-based caller authentication and identification algorithm for cellular communication networks that addresses these weaknesses. Voice data were collected from volunteer participants and the Mozilla Common Voice (MCV) database. Voice feature vectors were extracted from the voice data using the Mel Frequency Cepstral Coefficient (MFCC) technique. The extracted voice feature vectors were used to train a Multilayer Perceptron (MLP) neural network for voiceprint generation. The MLP neural network architecture was optimized through Neural Architecture Search (NAS) using the Neural Network Intelligence (NNI) toolkit. The optimized MLP neural network architecture was trained with the extracted voice feature vectors to generate a voiceprint generation model. The developed voiceprint model was then deployed to create an artificial neural network-based caller authentication and identification algorithm in cellular communication networks. The developed algorithm was evaluated in the cellular communication emulation setup using accuracy, false acceptance rate (FAR), and false rejection rate (FRR) as metrics. The results of the performance evaluation of the developed algorithm for authenticating first-time users of a new SIM and identifying third-party SIM users both showed 100% accuracy, 0% FAR, and 0% FRR. These results indicate that the developed algorithm has strong potential to prevent the use of black market SIMs and deter perpetrators from using third-party SIMs to access network services while remaining anonymous.