基于选择性超声波干扰的语音助手隐私保护系统  

A Privacy Protection System for Voice Assistants Based on Selective Ultrasonic Interference

作　　者：程健豪 周满 吴逸豪 CHENG Jianhao;ZHOU Man;WU Yihao(Key Laboratory of Aerospace Information Security and Trusted Computing,Ministry of Education,School of Cyber Science and Engineering,Wuhan University,Wuhan 430072,Hubei,China;School of Cyber Science and Engineering,Huazhong University of Science and Technology,Wuhan 430074,Hubei,China)

机构地区：[1]空天信息安全与可信计算教育部重点实验室,武汉大学国家网络安全学院,湖北武汉430072 [2]华中科技大学网络空间安全学院,湖北武汉430074

出　　处：《武汉大学学报（理学版）》2023年第5期553-564,共12页Journal of Wuhan University:Natural Science Edition

基　　金：国家自然科学基金(U20B2049)。

摘　　要：智能语音助手所要求的开放式麦克风权限带来了用户私密谈话内容泄露的风险。选择性超声波麦克风干扰系统是一种保护用户语音隐私安全的解决方案。然而,目前的系统无法有效地保护语音隐私安全,且去干扰机制的适用性存在问题。因此,本文提出了一种安全性更高、适用性更强的选择性超声波麦克风干扰系统。安全性方面,使用混合语音类型干扰可靠的混淆非授权设备;适用性方面,设计了一种参考高频段投影干扰副本的迭代自适应滤波方法,允许授权设备在多种干扰信号类型下记录更高质量的干净语音。此外,本文还拓展了选择性干扰机制,保留用户与语音助手的无接触交互能力。实验表明,本文设计的系统能够可靠地干扰非授权麦克风,同时授权设备去干扰效果在多项语音评价指标上均优于现有的选择性超声波麦克风干扰系统。The open microphone permission required by smart voice assistants brings the risk of user private conversation content leakage.The use of the selective ultrasonic microphone jamming system has been proposed as a means to protect user voice privacy and security.However,the current system is not fully effective in preserving speech privacy,and there are limitations with the active interference cancellation mechanism.Hence,this paper presents a selective ultrasonic microphone jamming system that addresses these challenges by incorporating enhanced safety considerations and improved applicability.This paper employs mixed speech type interference to effectively obfuscate unauthorized devices,enhancing security.Moreover,for improved applicability,an iterative adaptive filtering approach is designed,utilizing a reference high-frequency projection interference copy to enable authorized devices to capture cleaner speech amidst various types of interference signals.Furthermore,this paper extends the selective interference mechanism to preserve the touchless interaction capability between users and voice assistants.Experimental results demonstrate that the proposed system effectively blocks unauthorized microphones,and the quality of the denoised speech is higher than those recovered from existing selective ultrasonic jmicrophone amming systems in terms of subjective and objective speech evaluation.

关 键 词：系统安全 超声波麦克风干扰器 智能语音助手 隐私保护 主动干扰消除 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]