Anonymous networks intended to enhance privacy and evade censorship are also being exploited for abusive activities. Technical schemes have been proposed to selectively revoke the anonymity of abusive users, or simply limit them from anonymously accessing online service providers. We designed an empirical survey study to assess the effects of deploying these schemes on 75 users of the Tor anonymous network. We evaluated proposed schemes based on examples of the intended or abusive use cases they may address, their technical implementation and the types of entities responsible for enforcing them. Our results show that revocable anonymity schemes would particularly deter the intended uses of anonymous networks. We found a lower reported decrease in usage for schemes addressing spam than those directly compromising free expression. However, participants were concerned that all technical mechanisms for addressing anonymous abuses could be exploited beyond their intended goals (51.7%) to harm users (43.8%). Participants were distrustful of the enforcing entities involved (43.8%) and concerned about being unable to verify (49.3%) how particular mechanisms were applied.

New consent management platforms (CMPs) have been introduced to the web to conform with the EU's General Data Protection Regulation, particularly its requirements for consent when companies collect and process users' personal data. This work analyses how the most prevalent CMP designs affect people's consent choices. We scraped the designs of the five most popular CMPs on the top 10,000 websites in the UK (n=680). We found that dark patterns and implied consent are ubiquitous; only 11.8% meet our minimal requirements based on European law. Second, we conducted a field experiment with 40 participants to investigate how the eight most common designs affect consent choices. We found that notification style (banner or barrier) has no effect; removing the opt-out button from the first page increases consent by 22-23 percentage points; and providing more granular controls on the first page decreases consent by 8-20 percentage points. This study provides an empirical basis for the necessary regulatory action to enforce the GDPR, in particular the possibility of focusing on the centralised, third-party CMP services as an effective way to increase compliance.

We investigate whether post-hoc model explanations are effective for diagnosing model errors--model debugging. In response to the challenge of explaining a model's prediction, a vast array of explanation methods have been proposed. Despite increasing use, it is unclear if they are effective. To start, we categorize \textit{bugs}, based on their source, into: ~\textit{data, model, and test-time} contamination bugs. For several explanation methods, we assess their ability to: detect spurious correlation artifacts (data contamination), diagnose mislabeled training examples (data contamination), differentiate between a (partially) re-initialized model and a trained one (model contamination), and detect out-of-distribution inputs (test-time contamination). We find that the methods tested are able to diagnose a spurious background bug, but not conclusively identify mislabeled training examples. In addition, a class of methods, that modify the back-propagation algorithm are invariant to the higher layer parameters of a deep network; hence, ineffective for diagnosing model contamination. We complement our analysis with a human subject study, and find that subjects fail to identify defective models using attributions, but instead rely, primarily, on model predictions. Taken together, our results provide guidance for practitioners and researchers turning to explanations as tools for model debugging.

Most users of smartphone apps remain unaware of what data about them is being collected, by whom, and how these data are being used. In this mixed methods investigation, we examine the question of whether revealing key data collection practices of smartphone apps may help people make more informed privacy-related decisions. To investigate this question, we designed and prototyped a new class of privacy indicators, called Data Controller Indicators (DCIs), that expose previously hidden information flows out of the apps. Our lab study of DCIs suggests that such indicators do support people in making more confident and consistent choices, informed by a more diverse range of factors, including the number and nature of third-party companies that access users' data. Furthermore, personalised DCIs, which are contextualised against the other apps an individual already uses, enable them to reason effectively about the differential impacts on their overall information exposure.

Abstract Scatterplots have been in use for about two centuries, primarily for observing the relationship between two variables and commonly for supporting correlation analysis. In this paper, we report an empirical study that examines how humans’ perception of correlation using scatterplots relates to the Pearson's product‐moment correlation coefficient (PPMCC) – a commonly used statistical measure of correlation. In particular, we study human participants’ estimation of correlation under different conditions, e.g., different PPMCC values, different densities of data points, different levels of symmetry of data enclosures, and different patterns of data distribution. As the participants were instructed to estimate the PPMCC of each stimulus scatterplot, the difference between the estimated and actual PPMCC is referred to as an offset. The results of the study show that varying PPMCC values, symmetry of data enclosure, or data distribution does have an impact on the average offsets, while only large variations in density cause an impact that is statistically significant. This study indicates that humans’ perception of correlation using scatterplots does not correlate with computed PPMCC in a consistent manner. The magnitude of offsets may be affected not only by the difference between individuals, but also by geometric features of data enclosures. It suggests that visualizing scatterplots does not provide adequate support to the task of retrieving their corresponding PPMCC indicators, while the underlying model of humans’ perception of correlation using scatterplots ought to feature other variables in addition to PPMCC. The paper also includes a theoretical discussion on the cost‐benefit of using scatterplots.

This research measures human performance in inferring the functional types (i.e., home, work, leisure and transport) of locations in geo-location data using different visual representations of the data (textual, static and animated visualizations) along with different amounts of data (1, 3 or 5 day(s)). We first collected real life geo-location data from tweets. We then asked the data owners to tag their location points, resulting in ground truth data. Using this dataset we conducted an empirical study involving 45 participants to analyze how accurately they could infer the functional location of the original data owners under different conditions, i.e., three data representations, three data densities and four location types. The study results indicate that while visual techniques perform better than textual ones, the functional locations of human activities can be inferred with a relatively high accuracy even using only textual representations and a low density of location points. Workplace was more easily inferred than home while transport was the functional location with the highest accuracy. Our results also showed that it was easier to infer functional locations from data exhibiting more stable and consistent mobility patterns, which are thus more vulnerable to privacy disclosures. We discuss the implications of our findings in the context of privacy preservation and provide guidelines to users and companies to help preserve and safeguard people's privacy.

The aim of this research was to understand what affects people's privacy preferences in smartphone apps. We ran a four-week study in the wild with 34 participants. Participants were asked to answer questions, which were used to gather their personal context and to measure their privacy preferences by varying app name and purpose of data collection. Our results show that participants shared the most when no information about data access or purpose was given, and shared the least when both of these details were specified. When just one of either purpose or the requesting app was shown, participants shared less when just the purpose was specified than when just the app name was given. We found that the purpose for data access was the predominant factor affecting users' choices. In our study the purpose condition vary from being not specified, to vague to be very specific. Participants were more willing to disclose data when no purpose was specified. When a vague purpose was shown, participants became more privacy-aware and were less willing to disclose their information. When specific purposes were shown participants were more willing to disclose when the purpose for requesting the information appeared to be beneficial to them, and shared the least when the purpose for data access was solely beneficial to developers.

Many smartphone apps collect personal information used for a variety of purposes. Users, however, are often unaware of this kind of access even though they must grant the required permissions upon app installation. We have identified three reasons for this unawareness. First, relevant permissions c

We review current technical and social barriers to COPPA compliance for popular online services aimed at children. We show that complying with COPPA has proven difficult for developers, even when a genuine attempt was made. We investigate reasons for this lack of compliance and identify common causes: specifically, difficulties obtaining verifiable parental control as well as supply mechanisms for parents to understand, review, grant access and monitor collection of their children's personal data. Unless part of online services, mobile apps do not need to comply with COPPA.

Social apps usually require a lot of personal information in order to be tailored to the needs of individual users. However, the inherent social exchange of data exposes a user's personal data to other app users or publicly for anyone to see. In this paper, we present an app that enables users to determine the optimal location and time to meet without exposing their information to other users. We compare this app to other research-based and commercial social apps and show that ours is the only one where the risk of exposure is not present. In order to provide such improved privacy protections, we use openPDS, a decentralized and open-source framework. openPDS enables users to store their data on their own servers and participate in group computations without exposing their raw data.

Research on visualizing Semantic Web data has yielded many tools that rely on information visualization techniques to better support the user in understanding and editing these data. Most tools structure the visualization according to the concept definitions and interrelations that constitute the ontology’s vocabulary. Instances are often treated as somewhat peripheral information, when considered at all. These instances, that populate ontologies, represent an essential part of any knowledge base. Understanding instance-level data might be easier for users because of their higher concreteness, but instances will often be orders of magnitude more numerous than the concept definitions that give them machine-processable meaning. As such, the visualization of instance-level data poses different but real challenges. The authors present a visualization technique designed to enable users to visualize large instance sets and the relations that connect them. This visualization uses both node-link and adjacency matrix representations of graphs to visualize different parts of the data depending on their semantic and local structural properties. The technique was originally devised for simple social network visualization. The authors extend it to handle the richer and more complex graph structures of populated ontologies, exploiting ontological knowledge to drive the layout of, and navigation in, the representation embedded in a smooth zoomable environment.

Effective collaborative authoring techniques require tools that consider the social aspects of collaboration in addition to the technical aspects. Collaborative authoring is fundamentally different to individual writing because of the communications that must inevitably take place between team members.Despite the fact that collaborative authoring has greatly increased in popularity in recent years, most collaborative authoring efforts are performed using tools that are primarily designed for individual authors. The lack of regard for the collaborative process leads to a number of common problems.This paper presents research into the use of a prototype wiki-based system (CAWS) to more effectively support the collaborative process. The results of two field studies into the use of the system are examined, in order to investigate the effectiveness of the techniques employed by the tool.

This paper presents research into ways that e-learning can be integrated into conventional higher education teaching. It examines how appropriate techniques and technologies may be selected, adapted and combined for effectiveness, to accommodate the preferences or needs associated with disciplinary differences. The research reviews existing literature and surveys practice. Disciplinary differences are analyzed and examples of typical applications are presented with an analysis of those techniques which are most appropriate. It also analyses quantitative and qualitative data gathered from interviews with students across a number of academic disciplines. The conclusions have implications for the direction of future work on learning design

This poster presents research into ways in which electronic based methods can be used in teaching. Specifically it wishes to examine how appropriate techniques and technologies may be selected, tailored and combined for effectiveness, based on disciplinary differences.The research reviews existing literature, surveys existing practices and research into new techniques which may be adopted for electronic learning. It also analyses quantitative and qualitative data gathered from interviews with students across a number of academic disciplines.