Protecting voice assistants from threats via the acoustic channel

The study examines methods for protecting voice assistants from unauthorized control commands issued by attackers. The limitations of traditional filtering and verification techniques are identified, along with vulnerabilities and constraints of voice assistants. The effectiveness of existing and proposed protective measures is evaluated, including their applicability in typical scenarios. As part of the research, a method was proposed that combines software- and hardware-based security mechanisms for voice assistants, ensuring the protection of critical command execution.