| 引用本文： |
-
姜禹,施祺,丛洋,胡爱群.面向无线感知模型的用户隐私保护方法研究[J].信息安全学报,已采用 [点击复制]
- Jiang Yu,SHI Qi,CONG Yang,HU Aiqun.Research on Security and User Privacy Protection Meth-ods for Wireless Sensing Models[J].Journal of Cyber Security,Accept [点击复制]
|
|
| 摘要: |
| 作为通信感知一体化关键技术之一，基于Wi-Fi信道状态信息的无线感知技术展示了广泛的研究与应用前景，能够实现更细粒度、更高灵敏度的人体活动识别，适用于智能家居、移动健康监测等领域。然而，无线感知技术的应用场景往往会涉及到大量的个人隐私数据，随之带来了安全问题。攻击者能够利用无线感知技术，在未经用户同意的情况下追踪其活动习惯，从而侵犯用户隐私。因此，在设计无线感知系统时，需要考虑安全隐私保护的问题。本文对无线感知模型的用户隐私保护问题进行量化分析，通过对无线信号传播路径分解得到感知细粒度和感知信噪比两个量化因素，用于分析模型的感知能力以及感知范围。由于感知细粒度无法直接测量，本文提出一种基于多子载波复数比的量化因素提取方法，通过读取信道向量的幅度和角度变化来获得感知细粒度理论值。实验结果表明，基于多子载波复数比的量化因素提取方法可以更准确地识别人体动作，并提升动作感知的鲁棒性。利用该提取方法得到的感知细粒度准确反映了感知系统能够分辨的最小动作，感知信噪比准确反映了系统对某一动作的可感知范围。最后提出了无线感知模型的安全隐私风险评估算法以验证感知模型的安全性和隐私性，通过对不同级别用户提供针对性的安全引导，有效保障了感知场景的安全性与隐私性。 |
| 关键词: 隐私保护 无线感知 动作识别 子载波比 |
| DOI： |
| 投稿时间：2023-09-04修订日期：2024-03-03 |
| 基金项目:国家重点研发计划项目(2022YFB4300300) |
|
| Research on Security and User Privacy Protection Meth-ods for Wireless Sensing Models |
|
Jiang Yu1, SHI Qi2, CONG Yang2, HU Aiqun2
|
| (1.Southeast University;2.School of Cyber Science and Engineering, Southeast University, Nanjing 211189, China) |
| Abstract: |
| Within the landscape of integrated communication sensing, wireless sensing technology leveraging Wi-Fi channel state information has emerged as a pivotal tool with vast potential and extensive research and application prospects. Its capacity to discern finer nuances and exhibit heightened sensitivity in human activity recognition makes it a versatile asset across domains like smart homes and mobile health monitoring. However, the application scenarios of wireless sensing technology often involve vast amounts of personal privacy data, thus raising security concerns. Attackers can exploit wireless sensing technology to track users' activity patterns without their consent, thereby infringing upon user privacy. Therefore, when designing wireless sensing systems, the issues of security and privacy protection need to be considered. This paper quantitatively analyzes the user privacy protection issues of wireless sensing models. It decomposes the wire-less signal propagation path to obtain two quantitative factors: sensing granularity and sensing signal-to-noise ratio (SNR), used to analyze the model's sensing capability and sensing range. Since sensing granularity cannot be directly measured, this paper proposes a method for extracting quantitative factors based on the ratio of multi-subcarrier complex numbers. It obtains the theoretical value of sensing granularity by reading changes in the magnitude and angle of the channel vector. Experimental results demonstrate that the extraction method based on the ratio of multi-subcarrier complex numbers can more accurately identify human actions and enhance the robustness of action perception. The sensing granularity obtained using this method accurately reflects the smallest action that the sensing system can discern, while the sensing SNR accurately reflects the system's perceptible range for a particular action. Finally, a security and privacy risk assessment algorithm for wireless sensing models is proposed to validate the security and privacy of sensing models. By providing targeted security guidance for users of different levels, the security and privacy of sensing scenarios are effectively safeguarded. |
| Key words: Privacy Protection, Wireless Sensing, Action Recognition, Ratio of Subcarriers |
