Today, large amounts of valuable data are distributed among millions of user-held devices, such as personal computers, phones, or Internet-of-things devices. Many companies collect such data with the goal of using it for training machine learning models allowing them to improve their services. User-held data is, however, often sensitive, and collecting it is problematic in terms of privacy. We address this issue by proposing a novel way of training a supervised classifier in a distributed setting akin to the recently proposed federated learning paradigm, but under the stricter privacy requirement that the server that trains the model is assumed to be untrusted and potentially malicious. We thus preserve user privacy by design, rather than by trust. In particular, our framework, called secret vector machine (SecVM), provides an algorithm for training linear support vector machines (SVM) in a setting in which data-holding clients communicate with an untrusted server by exchanging messages designed to not reveal any personally identifiable information. We evaluate our model in two ways. First, in an offline evaluation, we train SecVM to predict user gender from tweets, showing that we can preserve user privacy without sacrificing classification performance. Second, we implement SecVM's distributed framework for the Cliqz web browser and deploy it for predicting user gender in a large-scale online evaluation with thousands of clients, outperforming baselines by a large margin and thus showcasing that SecVM is suitable for production environments.

Online tracking poses a serious privacy challenge that has drawn significant attention in both academia and industry. Existing approaches for preventing user tracking, based on curated blocklists, suffer from limited coverage and coarse-grained resolution for classification, rely on exceptions that impact sites' functionality and appearance, and require significant manual maintenance. In this paper we propose a novel approach, based on the concepts leveraged from $k$-Anonymity, in which users collectively identify unsafe data elements, which have the potential to identify uniquely an individual user, and remove them from requests.

The difficulty of partitioning social graphs has introduced new system design challenges for scaling of online social networks (OSNs). Vertical scaling by resorting to full replication can be a costly proposition. Scaling horizontally by partitioning and distributing data among multiple servers using, for e.g., distributed hash tables (DHTs), can suffer from expensive interserver communication. Such challenges have often caused costly rearchitecting efforts for popular OSNs like Twitter and Facebook. We design, implement, and evaluate SPAR, a Social Partitioning and Replication middleware that mediates transparently between the application and the database layer of an OSN. SPAR leverages the underlying social graph structure in order to minimize the required replication overhead for ensuring that users have their neighbors' data colocated in the same machine. The gains from this are multifold: Application developers can assume local semantics, i.e., develop as they would for a single machine; scalability is achieved by adding commodity machines with low memory and network I/O requirements; and N+K redundancy is achieved at a fraction of the cost. We provide a complete system design, extensive evaluation based on datasets from Twitter, Orkut, and Facebook, and a working implementation. We show that SPAR incurs minimum overhead, can help a well-known Twitter clone reach Twitter's scale without changing a line of its application logic, and achieves higher throughput than Cassandra, a popular key-value store database.

The mobile space is evolving at an astonishing rate with over 4.1 billion subscribers in existence. The world is also witnessing an explosion in social web services with more users seeking novel ways of interacting with friends and family. We are interested in the combination of these two exciting research spaces: the social web and the mobile web. We believe that the social mobile web is going to be a highly influential research area in the near future and given the huge growth that both these fields have experienced in recent times we feel that now is an excellent time to discuss this nascent research space. This workshop continues the successful social mobile web workshop held as part of SocialCom in 2009. The workshop explores the current state of the social mobile web and combines technical presentations, demos and position papers to drive interaction and discussion among participants.

The difficulty of scaling Online Social Networks (OSNs) has introduced new system design challenges that has often caused costly re-architecting for services like Twitter and Facebook. The complexity of interconnection of users in social networks has introduced new scalability challenges. Conventional vertical scaling by resorting to full replication can be a costly proposition. Horizontal scaling by partitioning and distributing data among multiples servers - e.g. using DHTs - can lead to costly inter-server communication.

Influenza can be transmitted through respirable (small airborne particles), inspirable (intermediate size), direct-droplet-spray, and contact modes. How these modes are affected by features of the virus strain (infectivity, survivability, transferability, or shedding profiles), host population (behavior, susceptibility, or shedding profiles), and environment (host density, surface area to volume ratios, or host movement patterns) have only recently come under investigation. A discrete-event, continuous-time, stochastic transmission model was constructed to analyze the environmental processes through which a virus passes from one person to another via different transmission modes, and explore which factors increase or decrease different modes of transmission. With the exception of the inspiratory route, each route on its own can cause high transmission in isolation of other modes. Mode-specific transmission was highly sensitive to parameter values. For example, droplet and respirable transmission usually required high host density, while the contact route had no such requirement. Depending on the specific context, one or more modes may be sufficient to cause high transmission, while in other contexts no transmission may result. Because of this, when making intervention decisions that involve blocking environmental pathways, generic recommendations applied indiscriminately may be ineffective; instead intervention choice should be contextualized, depending on the specific features of people, virus strain, or venue in question.

The typical state-of-the-art routing algorithms for delay tolerant networks are based on best next hop hill-climbing heuristics in order to achieve throughput and efficiency. The combination of these heuristics and the social network structure leads the routing to direct most of the traffic through a small subset of good users. For instance, in the SimBet algorithm, the top 10% of users carry out 54% of all the forwards and 85% of all the handovers. This unfair load distribution is not sustainable as it can quickly deplete constraint resources in heavily utilized mobile devices (e.g. storage, battery, budget, etc.). Moreover, because a small number of users carry a significant amount of the traffic, the system is not robust to random failures and attacks. To overcome these inefficiencies, this paper introduces Fair-Route, a routing algorithm for delay tolerant networks inspired by the social processes of perceived interaction strength, where messages are preferably forwarded to users that have a stronger social relation with the target of the message; and assortativity, that limits the exchange of messages to those users with similar "social status". We compare the performance of FairRoute to the state-of-the-art algorithms by extensive simulations on the MIT reality mining dataset. The results show that our algorithm outperforms existing algorithms in the de facto benchmark of throughput vs. forwards. Furthermore, it distributes better the load; the top 10% carry out 26% of the forwards and 28% of the handovers without any loss in performance.

A common approach to designing Recommender Systems (RS) consists of asking users to explicitly rate items in order to collect feedback about their preferences. However, users have been shown to be inconsistent and to introduce a non-negligible amount of natural noise in their ratings that affects the accuracy of the predictions. In this paper, we present a novel approach to improve RS accuracy by reducing the natural noise in the input data via a preprocessing step. In order to quantitatively understand the impact of natural noise, we first analyze the response of common recommendation algorithms to this noise. Next, we propose a novel algorithm to denoise existing datasets by means of re-rating: i.e. by asking users to rate previously rated items again. This denoising step yields very significant accuracy improvements. However, re-rating all items in the original dataset is unpractical. Therefore, we study the accuracy gains obtained when re-rating only some of the ratings.In particular, we propose two partial denoising strategies: data and user-dependent denoising. Finally, we compare the value of adding a rating of an unseen item vs. re-rating an item. We conclude with a proposal for RS to improve the quality of their user data and hence their accuracy: asking users to re-rate items might, in some circumstances, be more beneficial than asking users to rate unseen items.

Nearest-neighbor collaborative filtering provides a successful means of generating recommendations for web users. However, this approach suffers from several shortcomings, including data sparsity and noise, the cold-start problem, and scalability. In this work, we present a novel method for recommending items to users based on expert opinions. Our method is a variation of traditional collaborative filtering: rather than applying a nearest neighbor algorithm to the user-rating data, predictions are computed using a set of expert neighbors from an independent dataset, whose opinions are weighted according to their similarity to the user. This method promises to address some of the weaknesses in traditional collaborative filtering, while maintaining comparable accuracy. We validate our approach by predicting a subset of the Netflix data set. We use ratings crawled from a web portal of expert reviews, measuring results both in terms of prediction accuracy and recommendation list precision. Finally, we explore the ability of our method to generate useful recommendations, by reporting the results of a user-study where users prefer the recommendations generated by our approach.

Characterizing infectivity as a function of pathogen dose is integral to microbial risk assessment. Dose-response experiments usually administer doses to subjects at one time. Phenomenological models of the resulting data, such as the exponential and the Beta-Poisson models, ignore dose timing and assume independent risks from each pathogen. Real world exposure to pathogens, however, is a sequence of discrete events where concurrent or prior pathogen arrival affects the capacity of immune effectors to engage and kill newly arriving pathogens. We model immune effector and pathogen interactions during the period before infection becomes established in order to capture the dynamics generating dose timing effects. Model analysis reveals an inverse relationship between the time over which exposures accumulate and the risk of infection. Data from one time dose experiments will thus overestimate per pathogen infection risks of real world exposures. For instance, fitting our model to one time dosing data reveals a risk of 0.66 from 313 Cryptosporidium parvum pathogens. When the temporal exposure window is increased 100-fold using the same parameters fitted by our model to the one time dose data, the risk of infection is reduced to 0.09. Confirmation of this risk prediction requires data from experiments administering doses with different timings. Our model demonstrates that dose timing could markedly alter the risks generated by airborne versus fomite transmitted pathogens.