🤖 AI Summary
This paper addresses the design of differentially private algorithms for dynamic graphs under continual observation—where edges and nodes arrive and depart over time—and aims to release privacy-preserving statistical and optimization outputs in real time (e.g., triangle counting, minimum spanning tree, minimum cut, densest subgraph, maximum matching). We propose a unified framework integrating privacy-preserving accumulators, sensitivity clipping, and the exponential mechanism. Our approach achieves the first polylog(T) additive error for triangle counting; establishes a tight O(W log^{3/2} T / ε) additive error bound for minimum spanning tree; and proves an Ω(T) linear lower bound on additive error under user-level privacy. Across multiple problems, our algorithms simultaneously guarantee (1+β)-approximation accuracy and controllable additive error, bridging utility and privacy in evolving graph analytics.
📝 Abstract
Differentially private algorithms protect individuals in data analysis scenarios by ensuring that there is only a weak correlation between the existence of the user in the data and the result of the analysis. Dynamic graph algorithms maintain the solution to a problem (e.g., a matching) on an evolving input, i.e., a graph where nodes or edges are inserted or deleted over time. They output the value of the solution after each update operation, i.e., continuously. We study (event-level and user-level) differentially private algorithms for graph problems under continual observation, i.e., differentially private dynamic graph algorithms. We present event-level private algorithms for partially dynamic counting-based problems such as triangle count that improve the additive error by a polynomial factor (in the length $T$ of the update sequence) on the state of the art, resulting in the first algorithms with additive error polylogarithmic in $T$. We also give $varepsilon$-differentially private and partially dynamic algorithms for minimum spanning tree, minimum cut, densest subgraph, and maximum matching. The additive error of our improved MST algorithm is $O(W log^{3/2}T / varepsilon)$, where $W$ is the maximum weight of any edge, which, as we show, is tight up to a $(sqrt{log T} / varepsilon)$-factor. For the other problems, we present a partially-dynamic algorithm with multiplicative error $(1+eta)$ for any constant $eta>0$ and additive error $O(W log(nW) log(T) / (varepsilon eta))$. Finally, we show that the additive error for a broad class of dynamic graph algorithms with user-level privacy must be linear in the value of the output solution's range.
