Voice AI Systems Are Vulnerable to Hidden Audio Attacks
AI-powered voice and audio tools are becoming increasingly embedded in daily life, from digital assistants to smart speakers and customer service bots. Advances in large audio-language models (LALMs), which can both analyze and generate audio , now make it possible to control devices using voice commands, transcribe meetings automatically, or identify a song playing in the background. These models are also increasingly equipped with the ability to communicate with external se
New research reveals that advanced AI voice systems, including those used in digital assistants and smart speakers, are susceptible to hidden audio attacks. These attacks embed imperceptible sounds within audio clips, which can trick the AI into executing unauthorized commands without the user's awareness. The technique, called AudioHijack, has demonstrated a high success rate in manipulating leading AI models, forcing them to perform actions like sensitive web searches or downloading malicious files.
This vulnerability highlights a significant security risk in the widespread adoption of AI voice technologies, potentially exposing users to data breaches and unauthorized system control.
📌 Kaynak
Bu özet spectrumieee kaynağından otomatik derlenmiştir. Tamamı için orijinal habere gidin.
Orijinal haberi oku →