Practical Adversarial Attack Against Speech Recognition Platforms
Explore cutting-edge techniques in adversarial attacks against speech recognition platforms in this 58-minute conference talk by Shengzhi Zhang from Boston University's Department of Computer Science. Delve into the intricacies of exploiting vulnerabilities in popular voice-activated devices like Amazon Echo and Kodi. Gain insights into blackbox attack methodologies, philosophical implications, and intuitive approaches to compromising speech recognition systems. Witness live demonstrations showcasing the practical application of these techniques, and engage in a thought-provoking Q&A session. Conclude with an examination of local model vulnerabilities, equipping you with a comprehensive understanding of the current landscape in speech recognition security.