A Software Platform for Detecting Fraud in Audio Recordings Using Neural Network Models
Alexander V. Zhuravlev, Eugeny Mytarin, Vadim Moshkin · 2025
The article presents a software platform for automated fraud detection in audio recordings based on neural network models and ontological analysis. The platform integrates deep learning methods such as speech recognition (Vosk, Whisper), text lemmatization, intonation analysis, and matching with an ontological database to assess the likelihood of fraudulent actions. The solution is focused on applications in the field of financial security and countering social engineering. Special attention is paid to the flexibility of the system, including the ability to edit the ontology by the user. The results demonstrate the effectiveness of the platform in audio data processing, visualization of results, and integration with databases.