Many devices are accessed and controlled through voice assistants today, a representative example being Echo smart speakers and other Amazon devices controlled by Alexa. These offer the convenience of accessing services through voice interactions, but also raise privacy concerns, as data can be stored and used for personalization, and voice biometric information is sensitive. Unfortunately, there remains a lack of transparency and control over the collection and use of this data. Although prior work has shown evidence of ad targeting based on data derived from voice interactions and user profiles/interests, it has so far been an open question whether voice biometric information itself is utilized for targeting. In this paper, (i) we build a general auditing methodology to answer this question for off-the-shelf commercial smart speakers, and (ii) we apply it specifically to Amazon Echo Dot. Our findings suggest that Amazon Music ad content is more strongly associated with attributes (gender and age) related to voice characteristics than would be expected by chance. This has important implications for compliance, since voice contains sensitive biometric information that is protected by several privacy regulations.

Signal maps are essential for the planning and operation of cellular networks. However, the measurements needed to create such maps are expensive, often biased, not always reflecting the performance metrics of interest, and posing privacy risks. In this paper, we develop a unified framework for predicting cellular performance maps from limited available measurements. Our framework builds on a state-of-the-art random-forest predictor, or any other base predictor. We propose and combine three mechanisms that deal with the fact that not all measurements are equally important for a particular prediction task. First, we designquality-of-service functions ($Q$Q), including signal strength (RSRP) but also other metrics of interest to operators, such as number of bars, coverage (improving recall by 76%-92%) and call drop probability (reducing error by as much as 32%). By implicitly altering the loss function employed in learning, quality functions can also improve prediction for RSRP itself where it matters (e.g., MSE reduction up to 27% in the low signal strength regime, where high accuracy is critical). Second, we introduceweight functions($W$) to specify the relative importance of prediction at different locations and other parts of the feature space. We propose re-weighting based on importance sampling to obtain unbiased estimators when the sampling and target distributions are different. This yields improvements up to 20% for targets based on spatially uniform loss or losses based on user population density. Third, we apply theData Shapleyframework for the first time in this context: to assign values ($\phi$) to individual measurement points, which capture the importance of their contribution to the prediction task. This can improve prediction (e.g., from 64% to 94% in recall for coverage loss) by removing points with negative values and storing only the remaining data points (i.e., as low as 30%), which also has the side-benefit of helping privacy. We evaluate our methods and demonstrate significant improvement in prediction performance, using several real-world datasets.

We consider the problem of population density estimation based on location data crowdsourced from mobile devices, using kernel density estimation (KDE). In a conventional, centralized setting, KDE requires mobile users to upload their location data to a server, thus raising privacy concerns. Here, we propose a Federated KDE framework for estimating the user population density, which not only keeps location data on the devices but also provides probabilistic privacy guarantees against a malicious server that tries to infer users' location. Our approach Federated random Fourier feature (RFF) KDE leverages a random feature representation of the KDE solution, in which each user's information is irreversibly projected onto a small number of spatially delocalized basis functions, making precise localization impossible while still allowing population density estimation. We evaluate our method on both synthetic and real-world datasets, and we show that it achieves a better utility (estimation performance)-vs-privacy (distance between inferred and true locations) tradeoff, compared to state-of-the-art baselines (e.g., GeoInd). We also vary the number of basis functions per user, to further improve the privacy-utility trade-off, and we provide analytical bounds on localization as a function of areal unit size and kernel bandwidth.

Signal strength maps are of great importance to cellular providers for network planning and operation, however they are expensive to obtain and possibly limited or inaccurate in some locations. In this paper, we develop a prediction framework based on random forests to improve signal strength maps from limited measurements. First, we propose a random forests (RFs)-based predictor, with a rich set of features including location as well as time, cell ID, device hardware and other features. We show that our RFs-based predictor can significantly improve the tradeoff between prediction error and number of measurements needed compared to state-of-the-art data-driven predictors, i.e., requiring 80% less measurements for the same prediction accuracy, or reduces the relative error by 17% for the same number of measurements. Second, we leverage two types of real-world LTE RSRP datasets to evaluate into the performance of different prediction methods: (i) a small but dense Campus dataset, collected on a university campus and (ii) several large but sparser NYC and LA datasets, provided by a mobile data analytics company.

Community structure is an important property that captures inhomogeneities common in large networks, and modularity is one of the most widely used metrics for such community structure. In this paper, we introduce a principled methodology, the Spectral Graph Forge, for generating random graphs that preserves community structure from a real network of interest, in terms of modularity. Our approach leverages the fact that the spectral structure of matrix representations of a graph encodes global information about community structure. The Spectral Graph Forge uses a low-rank approximation of the modularity matrix to generate synthetic graphs that match a target modularity within user-selectable degree of accuracy, while allowing other aspects of structure to vary. We show that the Spectral Graph Forge outperforms state-of-the-art techniques in terms of accuracy in targeting the modularity and randomness of the realizations, while also preserving other local structural properties and node attributes. We discuss extensions of the Spectral Graph Forge to target other properties beyond modularity, and its applications to anonymization.

In this paper, we study the problem of generating synthetic graphs that resemble real-world graphs in terms of their degree correlations and potentially additional properties. We present an algorithmic framework that generates simple undirected graphs with the exact target joint degree matrix, which we refer to as 2K graphs, in linear time in the number of edges. Our framework imposes minimal constraints on the graph structure, which allows us to target additional graph properties during construction, namely, node attributes (2K+A), clustering (both average clustering, 2.25K, and degree-dependent clustering, 2.5K), and number of connected components (2K+CC). We also define, for the first time, the problem of directed 2K graph construction, provide necessary and sufficient conditions for realizability, and develop efficient construction algorithms. We evaluate our approach by creating synthetic graphs that target real-world graphs both undirected (such as Facebook) and directed (such as Twitter), and we show that it brings significant benefits, in terms of accuracy and running time, compared to the state-of-the-art approaches.

Community structure is an important property that captures inhomogeneities common in large networks, and modularity is one of the most widely used metrics for such community structure. In this paper, we introduce a principled methodology, the Spectral Graph Forge, for generating random graphs that preserves community structure from a real network of interest, in terms of modularity. Our approach leverages the fact that the spectral structure of matrix representations of a graph encodes global information about community structure. The Spectral Graph Forge uses a low-rank approximation of the modularity matrix to generate synthetic graphs that match a target modularity within user-selectable degree of accuracy, while allowing other aspects of structure to vary. We show that the Spectral Graph Forge outperforms state-of-the-art techniques in terms of accuracy in targeting the modularity and randomness of the realizations, while also preserving other local structural properties and node attributes. We discuss extensions of the Spectral Graph Forge to target other properties beyond modularity, and its applications to anonymization.

We study the problem of generating synthetic graphs that resemble real-world directed graphs in terms of their degree correlations. In order to capture degree correlation specifically for directed graphs, we define directed 2K (D2K) as those graphs with a given directed degree sequence (DDS) and a given target joint degree and attribute matrix (JDAM). We provide necessary and sufficient conditions for a target D2K to be realizable and we design an efficient algorithm that generates graph realizations with exactly the target D2K. We apply our algorithm to generate synthetic graphs that target real-world directed graphs (such as Twitter), and we demonstrate its benefits compared to state-of-the-art construction algorithms.

In this paper we demonstrate a novel approach to the use of spatio-temporally aggregated cell phone data to learn features of urban ecology (i.e., spatial distributions of distinct social and economic entities and their associated activities). Specifically, our technique involves four stages: (i) decomposing the aggregated cell phone activity within local areal units using spectral methods; (ii) learning spectral characteristics associated with ecological features using a training set; (iii) predicting local ecology composition for out-of-sample areas; and (iv) predicting activity time series for out-of-sample areas. The core of our approach is the projection of spectral features in cell phone activity series into an ecology-associated basis, allowing both identification of communication patterns arising from particular types of local activities and/or institutions and leveraging of those patterns for classification and activity prediction. We apply our methodology to aggregated communication and Internet traffic data from the cities of Milan and Trento to show the effectiveness of our method.

Scalable probabilistic modeling and prediction in high dimensional multivariate time-series, such as dynamic social networks with co-evolving nodes and edges, is a challenging problem, particularly for systems with hidden sources of dependence and/or homogeneity. Here, we address this problem through the discovery of hierarchical latent groups. We introduce a family of Conditional Latent Tree Models (CLTM), in which tree-structured latent variables incorporate the unknown groups. The latent tree itself is conditioned on observed covariates such as seasonality, historical activity, and node attributes. We propose a statistically efficient framework for learning both the hierarchical tree structure and the parameters of the CLTM. We demonstrate competitive performance on two real world datasets, one from the students' attempts at answering questions in a psychology MOOC and the other from Twitter users participating in an emergency management discussion and interacting with one another. In addition, our modeling framework provides valuable and interpretable information about the hidden group structures and their effect on the evolution of the time series.

The goal of this paper is to infer features of urban ecology (i.e., social and economic activities, and social interaction) from spatiotemporal cell phone activity data. We present a novel approach that consists of (i) time series decomposition of the aggregate cell phone activity per unit area using spectral methods, (ii) clustering of areal units with similar activity patterns, and (ii) external validation using a ground truth data set we collected from municipal and online sources. The key to our approach is the spectral decomposition of the original cell phone activity series into seasonal communication series (SCS) and residual communication series (RCS). The former captures regular patterns of socio-economic activity within an area and can be used to segment a city into distinct clusters. RCS across areas enables the detection of regions that are subject to mutual social influence and of regions that are in direct communication contact. The RCS and SCS thus provide distinct probes into the structure and dynamics of the urban environment, both of which can be obtained from the same underlying data. We illustrate the effectiveness of our methodology by applying it to aggregate Call Description Records (CDRs) from the city of Milan.

Several approaches have recently been proposed for modeling of continuous-time network data via dyadic event rates conditioned on the observed history of events and nodal or dyadic covariates. In many cases, however, interaction propensities – and even the underlying mechanisms of interaction – vary systematically across subgroups whose identities are unobserved. For static networks such heterogeneity has been treated via methods such as stochastic blockmodeling, which operate by assuming latent groups of individuals with similar tendencies in their group-wise interactions. Here we combine ideas from stochastic blockmodeling and continuous-time network models by positing a latent partition of the node set such that event dynamics within and between subsets evolve in potentially distinct ways. We illustrate the use of our model family by application to several forms of dyadic interaction data, including email communication and Twitter direct messages. Parameter estimates from the fitted models clearly reveal heterogeneity in the dynamics among groups of individuals. We also find that the fitted models have better predictive accuracy than both baseline models and relational event models that lack latent structure.

Our objective is to sample the node set of a large unknown graph via crawling, to accurately estimate a given metric of interest. We design a random walk on an appropriately defined weighted graph that achieves high efficiency by preferentially crawling those nodes and edges that convey greater information regarding the target metric. Our approach begins by employing the theory of stratification to find optimal node weights, for a given estimation problem, under an independence sampler. While optimal under independence sampling, these weights may be impractical under graph crawling due to constraints arising from the structure of the graph. Therefore, the edge weights for our random walk should be chosen so as to lead to an equilibrium distribution that strikes a balance between approximating the optimal weights under an independence sampler and achieving fast convergence. We propose a heuristic approach (stratified weighted random walk, or S-WRW) that achieves this goal, while using only limited information about the graph structure and the node properties. We evaluate our technique in simulation, and experimentally, by collecting a sample of Facebook college users. We show that S-WRW requires 13-15 times fewer samples than the simple re-weighted random walk (RW) to achieve the same estimation accuracy for a range of metrics.

State-of-the-art techniques for probability sampling of users of online social networks (OSNs) are based on random walks on a single social relation (typically friendship). While powerful, these methods rely on the social graph being fully connected. Furthermore, the mixing time of the sampling process strongly depends on the characteristics of this graph. In this paper, we observe that there often exist other relations between OSN users, such as membership in the same group or participation in the same event. We propose to exploit the graphs these relations induce, by performing a random walk on their union multigraph. We design a computationally efficient way to perform multigraph sampling by randomly selecting the graph on which to walk at each iteration. We demonstrate the benefits of our approach through (i) simulation in synthetic graphs, and (ii) measurements of Last.fm- an Internet website for music with social networking features. More specifically, we show that multigraph sampling can obtain a representative sample and faster convergence, even when the individual graphs fail, i.e., are disconnected or highly clustered.

Our goal in this paper is to develop a practical framework for obtaining a uniform sample of users in an online social network (OSN) by crawling its social graph. Such a sample allows to estimate any user property and some topological properties as well. To this end, first, we consider and compare several candidate crawling techniques. Two approaches that can produce approximately uniform samples are the Metropolis-Hasting random walk (MHRW) and a re-weighted random walk (RWRW). Both have pros and cons, which we demonstrate through a comparison to each other as well as to the "ground truth." In contrast, using Breadth-First-Search (BFS) or an unadjusted Random Walk (RW) leads to substantially biased results. Second, and in addition to offline performance assessment, we introduce online formal convergence diagnostics to assess sample quality during the data collection process. We show how these diagnostics can be used to effectively determine when a random walk sample is of adequate size and quality. Third, as a case study, we apply the above methods to Facebook and we collect the first, to the best of our knowledge, representative sample of Facebook users. We make it publicly available and employ it to characterize several key properties of Facebook.

With more than 250 million active users, Facebook (FB) is currently one of the most important online social networks. Our goal in this paper is to obtain a representative (unbiased) sample of Facebook users by crawling its social graph. In this quest, we consider and implement several candidate techniques. Two approaches that are found to perform well are the Metropolis-Hasting random walk (MHRW) and a re-weighted random walk (RWRW). Both have pros and cons, which we demonstrate through a comparison to each other as well as to the "ground-truth" (UNI - obtained through true uniform sampling of FB userIDs). In contrast, the traditional Breadth-First-Search (BFS) and Random Walk (RW) perform quite poorly, producing substantially biased results. In addition to offline performance assessment, we introduce online formal convergence diagnostics to assess sample quality during the data collection process. We show how these can be used to effectively determine when a random walk sample is of adequate size and quality for subsequent use (i.e., when it is safe to cease sampling). Using these methods, we collect the first, to the best of our knowledge, unbiased sample of Facebook. Finally, we use one of our representative datasets, collected through MHRW, to characterize several key properties of Facebook.

Situational awareness (SA) applications monitor the real world and the entities therein to support tasks such as rapid decision-making, reasoning, and analysis. Raw input about unfolding events may arrive from variety of sources in the form of sensor data, video streams, human observations, and so on, from which events of interest are extracted. Location is one of the most important attributes of events, useful for a variety of SA tasks. In this paper, we propose an approach to model and represent (potentially uncertain) event locations described by human reporters in the form of free text. We analyze several types of spatial queries of interest in SA applications. Our experimental evaluation demonstrates the effectiveness of our approach.

No abstract available.