恶魔音乐攻击智能语音系统.pdf

报告题目：“恶魔音乐”攻击智能语音系统 提纲 • 背景 • 相关工作 • 攻击方案 • 攻击原理 • 实验评估 • 攻击防御 背景——智能语音 背景——AI安全 • 传统攻击AI方法 恶意应用 病毒传播 无线劫持 重写固件 背景——AI安全 • 新型攻击AI方法 数据污染 窃取模型 背景——AI安全 • 新型攻击AI方法 数据污染 窃取模型 对抗样本 Stop + = Speed limit 50 [1] Ivan Evtimov, Kevin Eykholt, Earlence Fernandes, Tadayoshi Kohno, Bo Li, Atul Prakash, Amir Rahmati, and Dawn Song. Robust physical-world attacks on deep learning models. Computer Vision and Pattern Recognition, 2018. 提纲 • 背景 • 相关工作 • 攻击方案 • 攻击原理 • 实验评估 • 攻击防御 相关工作——智能语音系统攻击 [1] Nicholas Carlini, Pratyush Mishra, Tavish Vaidya, Yuankai Zhang, Micah Sherr, Clay Shields, David Wagner, and Wenchao Zhou. Hidden voice commands. In USENIX Security Symposium, pages 513–530, 2016. 相关工作——智能语音系统攻击 [1] Nicholas Carlini, Pratyush Mishra, Tavish Vaidya, Yuankai Zhang, Micah Sherr, Clay Shields, David Wagner, and Wenchao Zhou. Hidden voice commands. In USENIX Security Symposium, pages 513–530, 2016. [2] Guoming Zhang, Chen Yan, Xiaoyu Ji, Tianchen Zhang, Taimin Zhang, and Wenyuan Xu. Dolphinattack: Inaudible voice commands. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, pages 103–117. ACM, 2017. 相关工作——智能语音系统攻击 [1] Nicholas Carlini, Pratyush Mishra, Tavish Vaidya, Yuankai Zhang, Micah Sherr, Clay Shields, David Wagner, and Wenchao Zhou. Hidden voice commands. In USENIX Security Symposium, pages 513–530, 2016. [2] Guoming Zhang, Chen Yan, Xiaoyu Ji, Tianchen Zhang, Taimin Zhang, and Wenyuan Xu. Dolphinattack: Inaudible voice commands. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security, pages 103–117. ACM, 2017. [3] Nicholas Carlini