.png){kind=icon}
.png){kind=icon}
Return
Streaming Data Collection With a Private Sketch-Based Protocol
.png){kind=icon}
PRE
AI
DOI：10.1109/JIOT.2024.3397908.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Data stream collection is critical to analyze service conditions and detect anomalies in time, especially in Internet of Things. However, it may undermine the individual privacy. Local differential privacy (LDP) has recently become a popular privacy-preserving technique protecting users' privacy. However, most of them are still limited to the assumption of one-item collection, resulting in poor utility when extended to the multi-item collection from a very large domain. This article proposes a private streaming data collection framework, private sketch-based framework (PSF), which takes advantage of sketches. Combining the proposed background information and a decode-first collection-side workflow, the framework improves the utility by reducing the errors introduced by the sketching algorithm and the privacy budget utilization when collecting multiple items. We analytically prove the superior accuracy and privacy characteristics of PSF. In order to support specific computing tasks, we build two private protocols based on PSF, PrivSketch and PrivSketch+, aiming at frequency estimation and mean estimation, respectively. We demonstrate the utility of PrivSketch and PrivSketch+ theoretically, and also evaluate them experimentally. Our evaluation, with several diverse synthetic and real data sets, demonstrates that PrivSketch is 1-3 orders of magnitude better than the competitors in terms of utility in both frequency estimation and frequent item estimation, while being up to similar to 100x faster. PrivSketch+ performs similar to 4 orders of magnitude better than advanced solutions, such as piecewise mechanism (PM) and hybrid mechanism (HM), under a limited privacy budget.
Keywords:
Privacy
Estimation
Data collection
Frequency estimation
Protocols
Protection
Internet
local differential privacy (LDP)
mean estimation
sketch
Journal
IF:
8.9
Papers:
1.4W
Citations:
7.8W