MTSP-LDP: A Framework for Multi-Task Streaming Data Publication under Local Differential Privacy

TL;DR

MTSP-LDP tackles the problem of publishing useful analytics over infinite data streams under $w$-event LDP, addressing both utility degradation and the lack of multi-task support in prior methods. It introduces four interacting mechanisms: Private Dissimilarity Estimation to privately gauge current-versus-history changes, Optimal Privacy Budget Allocation (OBA) to distribute the privacy budget across a sliding window by exploiting temporal correlations, Private Adaptive Tree Publication with Data-Adaptive Tree Construction (ATC) plus cross-timestamp grouping and smoothing to produce accurate, time-consistent tree summaries, and a Budget-Free Multi-Task Streaming Query engine that uses released trees to answer counting, range, and event-monitoring queries without additional privacy cost. The framework provides formal $w$-event $\\epsilon$-LDP guarantees while achieving high utility on real-world datasets, outperforming state-of-the-art $w$-event LDP methods. The combination of data-adaptive trees and optimal budget allocation enables accurate multi-task streaming analytics without trusted servers, making MTSP-LDP practically impactful for privacy-preserving, real-time decision-making in domains like traffic and IoT.

Abstract

The proliferation of streaming data analytics in data-driven applications raises critical privacy concerns, as directly collecting user data may compromise personal privacy. Although existing $w$-event local differential privacy (LDP) mechanisms provide formal guarantees without relying on trusted third parties, their practical deployment is hindered by two key limitations. First, these methods are designed primarily for publishing simple statistics at each timestamp, making them inherently unsuitable for complex queries. Second, they handle data at each timestamp independently, failing to capture temporal correlations and consequently degrading the overall utility. To address these issues, we propose MTSP-LDP, a novel framework for \textbf{M}ulti-\textbf{T}ask \textbf{S}treaming data \textbf{P}ublication under $w$-event LDP. MTSP-LDP adopts an \emph{Optimal Privacy Budget Allocation} algorithm to dynamically allocate privacy budgets by analyzing temporal correlations within each window. It then constructs a \emph{data-adaptive private binary tree structure} to support complex queries, which is further refined by cross-timestamp grouping and smoothing operations to enhance estimation accuracy. Furthermore, a unified \emph{Budget-Free Multi-Task Processing} mechanism is introduced to support a variety of streaming queries without consuming additional privacy budget. Extensive experiments on real-world datasets demonstrate that MTSP-LDP consistently achieves high utility across various streaming tasks, significantly outperforming existing methods.

Figures (8)

• Figure 1: Representing user data as a private binary tree, using FO to estimate each node's property in LDP, and performing a range query on interval $[0,6]$.
• Figure 2: The framework of MTSP-LDP
• Figure 3: Illustration of OBA
• Figure 4: Illustration of ATC
• Figure 5: Example of a same tree without pruning (left) and pruning at node $a$ (right)

Theorems & Definitions (11)

• definition 1: $w$-neighboring
• definition 2: $w$-event LDP
• definition 3: Counting Query
• definition 4: Range Query
• definition 5: Event Monitoring
• theorem 1
• proof
• theorem 2
• proof
• theorem 3