Counterfactual (CF) explanations for machine learning (ML) models are preferred by end-users, as they explain the predictions of ML models by providing a recourse (or contrastive) case to individuals who are adversely impacted by predicted outcomes. Existing CF explanation methods generate recourses under the assumption that the underlying target ML model remains stationary over time. However, due to commonly occurring distributional shifts in training data, ML models constantly get updated in practice, which might render previously generated recourses invalid and diminish end-users trust in our algorithmic framework. To address this problem, we propose RoCourseNet, a training framework that jointly optimizes predictions and recourses that are robust to future data shifts. This work contains four key contributions: (1) We formulate the robust recourse generation problem as a tri-level optimization problem which consists of two sub-problems: (i) a bi-level problem that finds the worst-case adversarial shift in the training data, and (ii) an outer minimization problem to generate robust recourses against this worst-case shift. (2) We leverage adversarial training to solve this tri-level optimization problem by: (i) proposing a novel virtual data shift (VDS) algorithm to find worst-case shifted ML models via explicitly considering the worst-case data shift in the training dataset, and (ii) a block-wise coordinate descent procedure to optimize for prediction and corresponding robust recourses. (3) We evaluate RoCourseNet's performance on three real-world datasets, and show that RoCourseNet consistently achieves more than 96% robust validity and outperforms state-of-the-art baselines by at least 10% in generating robust CF explanations. (4) Finally, we generalize the RoCourseNet framework to accommodate any parametric post-hoc methods for improving robust validity.

Customizing the location obfuscation functions generated by existing systems can result in weakening the privacy guarantees offered by these functions as they are not robust against such updates. In this demo, we present a new framework called, CORGI, i.e., CustOmizable Robust Geo Indistinguishability. The demonstration platform is a web application which is built on top on a real world dataset (Gowalla). The user-friendly interface of the demo allows participants to easily specify their customization preferences and generate a customizable and robust location obfuscation function. They can also examine the trade-offs among privacy, utility, and customization; visualized on a map for comparison between CORGI and a state of the art baseline.

In this study, we present an in-depth analysis of users’ propensity toward negative and hateful behavior during the COVID-19 pandemic. We analyze a large dataset extracted from Twitter from the months of January 2020 up until June 2020. The dataset includes 2,470,888 tweets from 3,269 users who are active over a period of six months. We model users’ propensity toward hateful content over time by leveraging Random Forest regressor model and Long Short-Term Memory (LSTM) based many-to-one and Sequence2Sequence models for both short and long-term predictions. Our models leverage a set of features for each user, including the user’s psychological traits. We also study the impact of external triggers, such as COVID-related news concurrent with the users’ activities. To encode popular news, we propose using encoder states of a Sequence2Sequence model as features for a Tree-based regressor. The regressor, when combined with the vectorized news, results in an accurate prediction of tweeter’s hateful behavior in the short (decoder size of four weeks) and long term (decoder size of 10 weeks) with a total training data of 15 weeks x 3269 users. We also show that our model accurately profiles selected groups of users, as they are defined by specific psychological traits.

One of the most popular location privacy-preserving mechanisms applied in location-based services (LBS) is location obfuscation, where mobile users are allowed to report obfuscated locations instead of their real locations to services. Many existing obfuscation approaches consider mobile users that can move freely over a region. However, this is inadequate for protecting the location privacy of vehicles, as their mobility is restricted by external factors, such as road networks and traffic flows. This auxiliary information about external factors helps an attacker to shrink the search range of vehicles' locations, increasing the risk of location exposure.

In this article, we aim to answer an important set of questions about the potential longitudinal effects of repeated sharing and privacy settings decisions over jointly managed content among users in a social network. We model user interactions through a repeated game in a network graph. We present a variation of the one-shot Ultimatum Game, wherein individuals interact with peers to make a decision on a piece of shared content. The outcome of this game is either success or failure, wherein success implies that a satisfactory decision for all parties is made and failure instead implies that the parties could not reach an agreement. Our proposed game is grounded in empirical data about individual decisions in repeated pairwise negotiations about jointly managed content in a social network. We consider both a “continuous” privacy model as well the “discrete” case of a model wherein privacy values are to be chosen among a fixed set of options. We formally demonstrate that over time, the system converges toward a “fair” state, wherein each individual’s preferences are accounted for. Our discrete model is validated by way of a user study, where participants are asked to propose privacy settings for own shared content from a small, discrete set of options.

We use a dynamical systems perspective to analyze a collection of 2.4 million tweets known to originate from ISIS and ISIS-related users. From those users active over a long period of time (i.e., 2+ years), we derive sequences of behaviors and show that the top users cluster into behavioral classes, which naturally describe roles within the ISIS communication structure. We then correlate these classes to the retweet network of the top users showing the relationship between dynamic behavior and retweet network centrality. We use the underlying model to formulate informed hypotheses about the role each user plays. Finally, we show that this model can be used to detect outliers, i.e. accounts that are thought to be outside the ISIS organization but seem to be playing a key communications role and have dynamic behavior consistent with ISIS members.

As online image sharing has become commonplace, researchers have acknowledged the need to assist users in detecting sensitive (or private) images. However, image privacy classification tasks have shown to be nontrivial, as the designation of an image sensitivity requires considerations of the visual concepts in the image. In this paper, we propose an innovative framework that combines the power of knowledge transfer for efficient, personalized learning of individuals’ privacy preferences toward images.Our approach defines a meta-model, which, given the query image and a small set of labeled images (used for the user-privacy customization), identifies if the query image is private for a target user. A generic user can efficiently customize this model by providing a small labeled training set. Moreover, our proposed framework includes transfer learning techniques to import basic patterns for image processing learned from other domains. Transfer learning enables fast and accurate processing of images, and allows few shot learning to focus on customization. This helps speed up the training process and avoid risk of overfitting. Our proposed framework significantly outperforms several baselines, including advanced object-oriented approaches and other CNN-based methods.

User engagement in online public discourse often includes self-disclosure - the revelation of personal information. Such disclosures on online public platforms (e.g., news forums) become a shared history, vulnerable to detrimental use by advertisers and malicious parties. Yet, users indulge in self-disclosing behavior to attain strategic goals like relational development, social connectedness, identity clarification, and social control. In this work, we develop supervised models to detect instances of self-disclosure in users' comments in the context of public discourse. Using three different datasets, we validate the performance of our models. Our detection models achieve an accuracy of 75.8 percent in a news discourse dataset. The performances on evaluation against existing methods on two secondary datasets are on par if not better. We examine the rate at which users self-disclose to understand when and to what extent users abide by group norms of such behavior. Our results show that self-disclosing users are often similar in their alignment or divergence with the group norm. As such, these similarly divergent users in a conversation use similar language in their disclosures. Finally, we reflect o n the implications o f alignment with or divergence from group norms in light of online privacy.

To promote cost-effective task assignment in Spatial Crowdsourcing (SC), workers are required to report their location to servers, which raises serious privacy concerns. As a solution, geo-obfuscation has been widely used to protect the location privacy of SC workers, where workers are allowed to report perturbed location instead of the true location. Yet, most existing geo-obfuscation methods consider workers? mobility on a 2 dimensional (2D) plane, wherein workers can move in arbitrary directions. Unfortunately, 2D-based geo-obfuscation is likely to generate high traveling cost for task assignment over roads, as it cannot accurately estimate the traveling costs distortion caused by location obfuscation. In this paper, we tackle the SC worker location privacy problem over road networks. Considering the network-constrained mobility features of workers, we describe workers? mobility by a weighted directed graph, which considers the dynamic traffic condition and road network topology. Based on the graph model, we design a geo-obfuscation (GO) function for workers to maximize the workers? overall location privacy without compromising the task assignment efficiency. We formulate the problem of deriving the optimal GO function as a linear programming (LP) problem. By using the angular block structure of the LP's constraint matrix, we apply Dantzig-Wolfe decomposition to improve the time-efficiency of the GO function generation. Our experimental results in the real-trace driven simulation and the real-world experiment demonstrate the effectiveness of our approach in terms of both privacy and task assignment efficiency.

Machine labeling of image content as private or public is a notoriously difficult problem, with the usual image processing challenges compounded by the highly personal, subjective, and contextual nature of access control decision making. In general, a user's privacy expectation for a given image is consequential to specific contents therein and the presence of sensitive content somewhere in the image is sufficient to warrant a private label. In this work, we extend the problem of determining a single privacy label for a given image to jointly inferring a privacy label and detecting the specific areas of sensitive content within a privately labeled image. We propose a stochastic spatial attribution model which exploits sophisticated (deep neural net derived) image features over randomly selected image patches, as well as image saliency quantification. We validate our detected private regions through extensive user study experiments. This effort to achieve spatial attribution of private image content helps to lay a foundation for warning mechanisms which may serve to aid both social media sites and their users.

Crowdsourcing leverages the diverse skill sets of large collections of individual contributors to solve problems and execute projects, where contributors may vary significantly in experience, expertise, and interest in completing tasks. Hence, to ensure the satisfaction of its task requesters, most existing crowdsourcing platforms focus primarily on supervising contributors' behavior. This lopsided approach to supervision negatively impacts contributor engagement and platform sustainability.

Online users engage in self-disclosure - revealing personal information to others - in pursuit of social rewards. However, there are associated costs of disclosure to users' privacy. User profiling techniques support the use of contributed content for a number of purposes, e.g., micro-targeting advertisements. In this paper, we study self-disclosure as it occurs in newspaper comment forums. We explore a longitudinal dataset of about 60,000 comments on 2202 news articles from four major English news websites. We start with detection of language indicative of various types of self-disclosure, leveraging both syntactic and semantic information present in texts. Specifically, we use dependency parsing for subject, verb, and object extraction from sentences, in conjunction with named entity recognition to extract linguistic indicators of self-disclosure. We then use these indicators to examine the effects of anonymity and topic of discussion on self-disclosure. We find that anonymous users are more likely to self-disclose than identifiable users, and that self-disclosure varies across topics of discussion. Finally, we discuss the implications of our findings for user privacy.

In an effort to support users' decision making process in regards to shared and co-managed online images, in this paper we present a novel model to early detect images which may be subject to possible conflicting access control decisions. We present a group-based stochastic model able to identify potential privacy conflicts among multiple stakeholders of an image. We discuss experiments on a dataset of over 3000 online images, and compare our results with several baselines. Our approach outperforms all baselines, even the strong ones based on a Convolutional Neural Network architecture.

We propose a system called TwoFace to uncover crowdsourced review manipulators who target online review systems. A unique feature of TwoFace is its three-phase framework:(i) in the first phase, we intelligently sample actual evidence of manipulation(e.g., review manipulators) by exploiting low moderation crowdsourcing platforms that reveal evidence of strategic manipulation;(ii) we then propagate the suspiciousness of these seed users to identify similar users through a random walk over a "suspiciousness»» graph; and(iii) finally, we uncover(hidden) distant users who serve structurally similar roles by mapping users into a low-dimensional embedding space that captures community structure. Altogether, the TwoFace system recovers 83% to 93% of all manipulators in a sample from Amazon of 38,590 reviewers, even when the system is seeded with only a few samples from malicious crowdsourcing sites.

Online photo sharing is an increasingly popular activity for Internet users. More and more users are now constantly sharing their images in various social media, from social networking sites to online communities, blogs, and content sharing sites. In this article, we present an extensive study exploring privacy and sharing needs of users’ uploaded images. We develop learning models to estimate adequate privacy settings for newly uploaded images, based on carefully selected image-specific features. Our study investigates both visual and textual features of images for privacy classification. We consider both basic image-specific features, commonly used for image processing, as well as more sophisticated and abstract visual features. Additionally, we include a visual representation of the sentiment evoked by images. To our knowledge, sentiment has never been used in the context of image classification for privacy purposes. We identify the smallest set of features, that by themselves or combined together with others, can perform well in properly predicting the degree of sensitivity of users’ images. We consider both the case of binary privacy settings (i.e., public, private), as well as the case of more complex privacy options, characterized by multiple sharing options. Our results show that with few carefully selected features, one may achieve high accuracy, especially when high-quality tags are available.

Modeling, understanding, and predicting the spatio-temporal dynamics of online memes are important tasks, with ramifications on location-based services, social media search, targeted advertising and content delivery networks. However, the raw data revealing these dynamics are often incomplete and error-prone; for example, API limitations and data sampling policies can lead to an incomplete (and often biased) perspective on these dynamics. Hence, in this paper, we investigate new methods for uncovering the full (underlying) distribution through a novel spatio-temporal dynamics recovery framework which models the latent relationships among locations, memes, and times. By integrating these hidden relationships into a tensor-based recovery framework -- called AirCP -- we find that high-quality models of meme spread can be built with access to only a fraction of the full data. Experimental results on both synthetic and real-world Twitter hashtag data demonstrate the promising performance of the proposed framework: an average improvement of over 27% in recovering the spatio-temporal dynamics of hashtags versus five state-of-the-art alternatives.

Crowdsourcing allows many people to complete tasks of various difficulty with minimal recruitment and administration costs. However, the lack of participant accountability may entice people to complete as many tasks as possible without fully engaging in them, jeopardizing the quality of responses. In this paper, we present a dynamic and time efficient solution to the task assignment problem in crowdsourcing platforms. Our proposed approach, CrowdSelect, offers a theoretically proven algorithm to assign workers to tasks in a cost efficient manner, while ensuring high accuracy of the overall task. In contrast to existing works, our approach makes minimal assumptions on the probability of error for workers, and completely removes the assumptions that such probability is known apriori and that it remains consistent over time. Through experiments over real Amazon Mechanical Turk traces and synthetic data, we find that CrowdSelect has a significant gain in term of accuracy compared to state-of-the-art algorithms, and can provide a 17.5\% gain in answers' accuracy compared to previous methods, even when there are over 50\% malicious workers.

In this paper, we describe a novel approach to investigate negative behavior dynamics in online social networks as epidemic phenomena. We present a finite-state machine model for time-varying epidemic dynamics, and validate this model with experiments over a large dataset of Youtube commentaries, indicating how different epidemic patterns of behavior can be tied to specific interaction patterns among users. A full version of this paper is available on arXiv.org.

Cyberbullying is an increasingly prevalent phenomenon impacting young adults. In this paper, we present a study on both detecting cyberbullies in online social networks and identifying the pairwise interactions between users through which the influence of bullies seems to spread. In particular, we investigate the role of user demographics and social network features in predicting how users will respond to a cyberbullying comment. We characterize the influencer/influenced relationship by which a user who has no history of abuse observes a peer engaging in bullying and follows suit. To our knowledge, this is the first effort modeling peer pressure and social dynamics with analytical models. We validate our models on two distinct social network datasets, totalling over 16,000 posts. Our results offer insight into the dynamics of bullying and confirm social theories on the power of peer groups in the cyberworld. A full version of this paper is available on arXiv.org.

Online reviews are a cornerstone of consumer decision making. However, their authenticity and quality has proven hard to control, especially as polluters target these reviews toward promoting products or in degrading competitors. In a troubling direction, the widespread growth of crowdsourcing platforms like Mechanical Turk has created a large-scale, potentially difficult-to-detect workforce of malicious review writers. Hence, this paper tackles the challenge of uncovering crowdsourced manipulation of online reviews through a three-part effort: (i) First, we propose a novel sampling method for identifying products that have been targeted for manipulation and a seed set of deceptive reviewers who have been enlisted through crowdsourcing platforms. (ii) Second, we augment this base set of deceptive reviewers through a reviewer-reviewer graph clustering approach based on a Markov Random Field where we define individual potentials (of single reviewers) and pair potentials (between two reviewers). (iii) Finally, we embed the results of this probabilistic model into a classification framework for detecting crowd-manipulated reviews. We find that the proposed approach achieves up to 0.96 AUC, outperforming both traditional detection methods and a SimRank-based alternative clustering approach.

With the increasing volume of images users share through social sites, maintaining privacy has become a major problem, as demonstrated by a recent wave of publicized incidents where users inadvertently shared personal information. In light of these incidents, the need of tools to help users control access to their shared content is apparent. Toward addressing this need, we propose an Adaptive Privacy Policy Prediction (A3P) system to help users compose privacy settings for their images. We examine the role of social context, image content, and metadata as possible indicators of users' privacy preferences. We propose a two-level framework which according to the user's available history on the site, determines the best available privacy policy for the user's images being uploaded. Our solution relies on an image classification framework for image categories which may be associated with similar policies, and on a policy prediction algorithm to automatically generate a policy for each newly uploaded image, also according to users' social features. Overtime, the generated policies will follow the evolution of users' privacy attitude. We provide the results of our extensive evaluation over 5,000 policies, which demonstrate the effectiveness of our system, with prediction accuracies over 90 percent.

Abstract Online Social Networks (OSNs) facilitate the creation and maintenance of interpersonal online relationships. Unfortunately, the availability of personal data on social networks may unwittingly expose users to numerous privacy risks. As a result, establishing effective methods to control personal data and maintain privacy within these OSNs have become increasingly important. This research extends the current access control mechanisms employed by OSNs to protect private information shared among users of OSNs. The proposed approach presents a system of collaborative content management that relies on an extended notion of a “content stakeholder.” A tool, Collaborative Privacy Management (CoPE), is implemented as an application within a popular social‐networking site, facebook.com , to ensure the protection of shared images generated by users. We present a user study of our CoPE tool through a survey‐based study (n=80). The results demonstrate that regardless of whether Facebook users are worried about their privacy, they like the idea of collaborative privacy management and believe that a tool such as CoPE would be useful to manage their personal information shared within a social network.

Online social networking sites are experiencing tremendous user growth with hundreds of millions of active users. As a result, there is a tremendous amount of user profile data online, e.g., name, birthdate, etc. Protecting this data is a challenge. The task of access policy composition is a tedious and confusing effort for the average user having hundreds of friends. We propose an approach that assists users in composing and managing their access control policies. Our approach is based on a supervised learning mechanism that leverages user provided example policy settings as training sets to build classifiers that are the basis for auto-generated policies. Furthermore, we provide mechanisms to enable users to fuse policy decisions that are provided by their friends or others in the social network. These policies then regulate access to user profile objects. We implemented our approach and, through extensive experimentation, prove the accuracy of our proposed mechanisms.

Social Networking is one of the major technological phenomena of the Web 2.0, with hundreds of millions of people participating. Social networks enable a form of self expression for users, and help them to socialize and share content with other users. In spite of the fact that content sharing represents one of the prominent features of existing Social Network sites, Social Networks yet do not support any mechanism for collaborative management of privacy settings for shared content. In this paper, we model the problem of collaborative enforcement of privacy policies on shared data by using game theory. In particular, we propose a solution that offers automated ways to share images based on an extended notion of content ownership. Building upon the Clarke-Tax mechanism, we describe a simple mechanism that promotes truthfulness, and that rewards users who promote co-ownership. We integrate our design with inference techniques that free the users from the burden of manually selecting privacy preferences for each picture. To the best of our knowledge this is the first time such a protection mechanism for Social Networking has been proposed. In the paper, we also show a proof-of-concept application, which we implemented in the context of Facebook, one of today’s most popular social networks. We show that supporting these type of solutions is not also feasible, but can be implemented through a minimal increase in overhead to end-users.

A typical Denial of Service attack against a DBMS may occur through a query flood, that is, a large number of queries and/or updates sent by a malicious subject or several colluding malicious subjects to a target database with the intention to hinder other subjects from being serviced. In this paper we present experimental results showing that such attacks indeed degrade the performance of the DBMS; our experiments are conducted on several well known DBMS. We then propose some simple yet effective techniques for detecting query-flood attacks and protecting a DBMS against them.

We present Trust-/spl Xscr/;, a comprehensive XML-based framework for trust negotiations, specifically conceived for a peer-to-peer environment. Trust negotiation is a promising approach for establishing trust in open systems like the Internet, where sensitive interactions may often occur between entities at first contact, with no prior knowledge of each other. The framework we propose takes into account all aspects related to negotiations, from the specification of the profiles and policies of the involved parties to the selection of the best strategy to succeed in the negotiation. Trust-/spl Xscr/; presents a number of innovative features, such as the support for protection of sensitive policies, the use of trust tickets to speed up the negotiation, and the support of different strategies to carry on a negotiation. In this paper, besides presenting the language to encode security information, we present the system architecture and algorithms according to which negotiations take place.