Schloss Dagstuhl – Leibniz Center for Informatics
DROPS Dagstuhl Research Online Publication ServerNot a member yet
23028 research outputs found
Sort by
Research Infrastructures and Tools for Collaborative Networked Systems Research (Dagstuhl Seminar 24462)
No full textThis report presents the program and outcomes of Dagstuhl Seminar "Research Infrastructures and Tools for Collaborative Networked Systems Research" (24462). The seminar brought together experts from the network and distributed systems testbed community, scientists who rely on testbeds for their research, and representatives from funding agencies. It focused on bridging the gap between the services provided by large-scale testbed infrastructures and the needs of researchers conducting cutting-edge experiments. Discussions centered on enhancing the value and impact of research infrastructures by improving collaboration, streamlining experiment workflows, and developing testbed-agnostic tools. The goal was to make experimental research more modular, adaptable, and reproducible, ensuring that experiments and evaluation software can be easily modified, extended, and ported across different testbed environments. Key topics included strategies to improve research quality, reproducibility, and reusability, enhance the discovery process, and maximize the efficient use of research infrastructure resources
Private Estimation When Data and Privacy Demands Are Correlated
Get PDFDifferential Privacy (DP) is the current gold-standard for ensuring privacy for statistical queries. Estimation problems under DP constraints appearing in the literature have largely focused on providing equal privacy to all users. We consider the problems of empirical mean estimation for univariate data and frequency estimation for categorical data, both subject to heterogeneous privacy constraints. Each user, contributing a sample to the dataset, is allowed to have a different privacy demand. The dataset itself is assumed to be worst-case and we study both problems under two different formulations - first, where privacy demands and data may be correlated, and second, where correlations are weakened by random permutation of the dataset. We establish theoretical performance guarantees for our proposed algorithms, under both PAC error and mean-squared error. These performance guarantees translate to minimax optimality in several instances, and experiments confirm superior performance of our algorithms over other baseline techniques
Kernel Multiaccuracy
Get PDFPredefined demographic groups often overlook the subpopulations most impacted by model errors, leading to a growing emphasis on data-driven methods that pinpoint where models underperform. The emerging field of multi-group fairness addresses this by ensuring models perform well across a wide range of group-defining functions, rather than relying on fixed demographic categories. We demonstrate that recently introduced notions of multi-group fairness can be equivalently formulated as integral probability metrics (IPM). IPMs are the common information-theoretic tool that underlie definitions such as multiaccuracy, multicalibration, and outcome indistinguishably. For multiaccuracy, this connection leads to a simple, yet powerful procedure for achieving multiaccuracy with respect to an infinite-dimensional class of functions defined by a reproducing kernel Hilbert space (RKHS): first perform a kernel regression of a model’s errors, then subtract the resulting function from a model’s predictions. We combine these results to develop a post-processing method that improves multiaccuracy with respect to bounded-norm functions in an RKHS, enjoys provable performance guarantees, and, in binary classification benchmarks, achieves favorable multiaccuracy relative to competing methods
Count on Your Elders: Laplace vs Gaussian Noise
Get PDFIn recent years, Gaussian noise has become a popular tool in differentially private algorithms, often replacing Laplace noise which dominated the early literature on differential privacy. Gaussian noise is the standard approach to approximate differential privacy, often resulting in much higher utility than traditional (pure) differential privacy mechanisms. In this paper we argue that Laplace noise may in fact be preferable to Gaussian noise in many settings, in particular when we seek to achieve (ε,δ)-differential privacy for small values of δ. We consider two scenarios:
First, we consider the problem of counting under continual observation and present a new generalization of the binary tree mechanism that uses a k-ary number system with negative digits to improve the privacy-accuracy trade-off. Our mechanism uses Laplace noise and whenever δ is sufficiently small it improves the mean squared error over the best possible (ε,δ)-differentially private factorization mechanisms based on Gaussian noise. Specifically, using k = 19 we get an asymptotic improvement over the bound given in the work by Henzinger, Upadhyay and Upadhyay (SODA 2023) when δ = O(T^{-0.92}).
Second, we show that the noise added by the Gaussian mechanism can always be replaced by Laplace noise of comparable variance for the same (ε, δ)-differential privacy guarantee, and in fact for sufficiently small δ the variance of the Laplace noise becomes strictly better. This challenges the conventional wisdom that Gaussian noise should be used for high-dimensional noise.
Finally, we study whether counting under continual observation may be easier in an average-case sense than in a worst-case sense. We show that, under pure differential privacy, the expected worst-case error for a random input must be Ω(log(T)/ε), matching the known lower bound for worst-case inputs
Infinitely Divisible Noise for Differential Privacy: Nearly Optimal Error in the High ε Regime
Get PDFDifferential privacy (DP) can be achieved in a distributed manner, where multiple parties add independent noise such that their sum protects the overall dataset with DP. A common technique here is for each party to sample their noise from the decomposition of an infinitely divisible distribution. We analyze two mechanisms in this setting: 1) the generalized discrete Laplace (GDL) mechanism, whose distribution (which is closed under summation) follows from differences of i.i.d. negative binomial shares, and 2) the multi-scale discrete Laplace (MSDLap) mechanism, a novel mechanism following the sum of multiple i.i.d. discrete Laplace shares at different scales. For ε ≥ 1, our mechanisms can be parameterized to have O(Δ³ e^{-ε}) and O(min(Δ³ e^{-ε}, Δ² e^{-2ε/3})) MSE, respectively, where Δ denote the sensitivity; the latter bound matches known optimality results. Furthermore, the MSDLap mechanism has the optimal MSE including constants as ε → ∞. We also show a transformation from the discrete setting to the continuous setting, which allows us to transform both mechanisms to the continuous setting and thereby achieve the optimal O(Δ² e^{-2ε / 3}) MSE. To our knowledge, these are the first infinitely divisible additive noise mechanisms that achieve order-optimal MSE under pure DP for either the discrete or continuous setting, so our work shows formally there is no separation in utility when query-independent noise adding mechanisms are restricted to infinitely divisible noise. For the continuous setting, our result improves upon Pagh and Stausholm’s Arete distribution which gives an MSE of O(Δ² e^{-ε/4}) [Pagh and Stausholm, 2022]. Furthermore, we give an exact sampler tuned to efficiently implement the MSDLap mechanism, and we apply our results to improve a state of the art multi-message shuffle DP protocol from [Balle et al., 2020] in the high ε regime
Temporal Connectivity Augmentation
Get PDFConnectivity in temporal graphs relies on the notion of temporal paths, in which edges follow a chronological order (either strict or non-strict). In this work, we investigate the question of how to make a temporal graph connected. More precisely, we tackle the problem of finding, among a set of proposed temporal edges, the smallest subset such that its addition makes the graph temporally connected (TCA). We study the complexity of this problem and variants, under restricted lifespan of the graph, i.e. the maximum time step in the graph. Our main result on TCA is that for any fixed lifespan at least 2, it is NP-complete in both the strict and non-strict setting. We additionally provide a set of restrictions in the non-strict setting which makes the problem solvable in polynomial time and design an algorithm achieving this complexity. Interestingly, we prove that the source variant (making a given vertex a source in the augmented graph) is as difficult as TCA. On the opposite, we prove that the version where a list of connectivity demands has to be satisfied is solvable in polynomial time, when the size of the list is fixed. Finally, we highlight a variant of the previous case for which even with two pairs the problem is already NP-hard
Dismountability in Temporal Cliques Revisited
Get PDFA temporal graph is a graph whose edges are available only at certain points in time. It is temporally connected if the nodes can reach each other by paths that traverse the edges chronologically (temporal paths). Unlike static graphs, temporal graphs do not always admit small subsets of edges that preserve connectivity (temporal spanners) - there exist temporal graphs with Θ(n²) edges, all of which are critical. In the case of temporal cliques (the underlying graph is complete), spanners of size O(nlog n) are guaranteed. The original proof of this result by Casteigts et al. [ICALP 2019] combines a number of techniques, one of which is called dismountability. In a recent work, Angrick et al. [ESA 2024] simplified the proof and showed, among other things, that a one-sided version of dismountability can replace elegantly the second part of the proof.
In this paper, we revisit methodically the dismountability principle. We start by characterizing the structure that a temporal clique must have if it is non 1-hop dismountable, then neither 1-hop nor 2-hop (i.e. non {1,2}-hop) dismountable, and finally non {1,2,3}-hop dismountable. It turns out that if a clique is k-hop dismountable for any other k, then it must also be {1,2,3}-hop dismountable, thus no additional structure can be obtained beyond this point. Interestingly, excluding 1-hop and 2-hop dismountability is already sufficient for reducing the spanner problem from cliques to extremally matched bicliques, where the O(nlog n) result is subsequently obtained. Put together with the strategy of Angrick et al., this entire result can now be recovered using only dismountability. An interesting by-product of our analysis is that any minimal counter-example to the existence of 4n spanners must satisfy the properties of non {1,2,3}-hop dismountable cliques.
In the second part, we discuss further connections between dismountability and another technique called pivotability. In particular, we show that if a temporal clique is recursively k-hop dismountable, then it is also pivotable (and thus admits a 2n spanner, whatever k). We also study a family of labelings called full-range that forces both dismountability and pivotability. The latter gives some evidence that large lifetimes could be exploited more generally for the construction of spanners
Brief Announcement: Anonymous Distributed Localisation via Spatial Population Protocols
Get PDFIn the distributed localization problem (DLP), n anonymous robots (agents) A₀, …, A_{n-1} begin at arbitrary positions p₀, …, p_{n-1} ∈ S, where S is a Euclidean space. Initially, each agent A_i operates within its own coordinate system in S, which may be inconsistent with those of other agents. The primary goal in DLP is for agents to reach a consensus on a unified coordinate system that accurately reflects the relative positions of all points, p₀, …, p_{n-1}, in S. Extensive research on DLP has primarily focused on the feasibility and complexity of achieving consensus when agents have limited access to inter-agent distances, often due to missing or imprecise data. In this paper, however, we examine a minimalist, computationally efficient model of distributed computing in which agents have access to all pairwise distances, if needed. Specifically, we introduce a novel variant of population protocols, referred to as the spatial population protocols model. In this variant each agent can memorise one or a fixed number of coordinates, and when agents A_i and A_j interact, they can not only exchange their current knowledge but also either determine the distance d_{ij} between them in S (distance query model) or obtain the vector v→_{ij} spanning points p_i and p_j (vector query model). We present here a leader-based localisation protocol with distance queries