Транскрібація на малопотужних процесорах: аналіз технологій та загроз інформаційній безпеці

Abstract

The paper examines technologies for automatic audio transcription on low power processors and analyzes related information security threats. The study reviews microcontrollers and single-board computers used for local speech recognition and considers TinyML-based models suitable for devices with limited resources. Particular attention is paid to optimization methods that enable efficient transcription on constrained hardware. The research also identifies potential risks related to covert audio data collection and transmission. The results highlight the need for improved monitoring and security measures in systems using autonomous edge devices.