About
Devil's Whisper is a BrainHack 2021 Showcase to raise awareness on audio Adversarial Attacks on smartphones and smart home devices.
Disclaimer
The audio Adversarial Attack's success may depend on various factors, including but not limited to:
| Audio file type (wav, mp3) |
| Volume of audio |
| Speaker type and quality |
| Distance between speaker and target device |
| Background noise |
| Type of target device |
| Version of target device |
| Microphone sensitivity of the target |
References
Code for adversarial attack adapted from https://github.com/RiskySignal/Devil-Whisper-Attack
Chen, Y., Yuan, X., Zhang, J., Zhao, Y., Zhang, S., Chen, K., & Wang, X. (2020). Devil's Whisper: A General Approach for Physical Adversarial Attacks against Commercial Black-box Speech Recognition Devices. USENIX Security Symposium.
GIF of Adversarial Image Patch obtained from https://www.theverge.com/2018/1/3/16844842/ai-computer-vision-trick-adversarial-patches-google.
Brown, T. B., Mané, D., Roy, A., Abadi, M., & Gilmer, J. (2017). Adversarial patch. arXiv preprint arXiv:1712.09665.
Attributions
Icons
- Hacker Icon made by Freepik from www.flaticon.com
- All other Icons from Fontawesome
Photos
- Whatsapp from Pexels by Alok Sharma
- Google Home from Unsplash by Thomas Kolnowski
Audio Waveform
- Audio Waveforms from Wavesurfer JS