In this paper, we present a novel method to predict the long-term popularity of user-generated content (UGC). At first, the method clusters the dynamics of UGC popularity into a vocabulary of growth in popularity (sequence) by using a mixture model. Eventually, the method assigns to each sequence a topic model to describe the dynamics of the sequence in a compact way. We then use this topic model to identify similar patterns of growth in popularity of newly observed UGC. The proposed method has two key features: First, it considers the historical dynamics of the UGC popularity, and second it provides long-term popularity prediction. Results on the real dataset of UGC show that the proposed method is flexible, and able to accurately forecast the complete growth in popularity of a given UGC.

Ad networks use the behaviors of online users to associate them with preferences (features), and market these features to enable advertisers to target online users. Typical features associated with users include location, interests, gender, age, and etc. Furthemore, ad networks provide their clients with campaing creation tools to help to them to configure and run campains. In this paper, we study the pricing of ads using the ad campaing planning tools of ad networks. We develop tools to collect the suggested bid prices from two platforms: YouTube and Facebook. Analyzing these prices we find that United States is the most expensive country in both platforms. We also find that the most expensive preferences are different in YouTube and Facebook. In YouTube, the top preferences are related to Oil & Gas, while in Facebook are devices, ethnics or politics depending on the type of bidding. Finally, we do not find any price difference genders in Facebook.

Numerous important problems can be framed as learning from graph data. We propose a framework for learning convolutional neural networks for arbitrary graphs. These graphs may be undirected, directed, and with both discrete and continuous node and edge attributes. Analogous to image-based convolutional networks that operate on locally connected regions of the input, we present a general approach to extracting locally connected regions from graphs. Using established benchmark data sets, we demonstrate that the learned feature representations are competitive with state of the art graph kernels and that their computation is highly efficient.

While substantial effort has been devoted to understand fraudulent activity in traditional online advertising (search and banner), more recent forms such as video ads have received little attention. The understanding and identification of fraudulent activity (i.e., fake views) in video ads for advertisers, is complicated as they rely exclusively on the detection mechanisms deployed by video hosting portals. In this context, the development of independent tools able to monitor and audit the fidelity of these systems are missing today and needed by both industry and regulators.

Similarity computation between pairs of objects is often a bottleneck in many applications that have to deal with massive volumes of data. Motivated by applications such as collaborative filtering in large-scale recommender systems, and influence probabilities learning in social networks, we present new randomized algorithms for the estimation of weighted similarity in data streams.

Network Function Virtualization is pushing network operators to deploy commodity hardware that will be used to run middlebox functionality and processing on behalf of third parties: in effect, network operators are slowly but surely becoming in-network cloud providers. The market for innetwork clouds is large, ranging from content providers, mobile applications and even end-users.

To assess the performance of caching systems, the definition of a proper process describing the content requests generated by users is required. Starting from the analysis of traces of YouTube video requests collected inside operational networks, we identify the characteristics of real traffic that need to be represented and those that instead can be safely neglected. Based on our observations, we introduce a simple, parsimonious traffic model, named shot noise model (SNM), that allows us to capture temporal and geographical locality of content popularity. The SNM is sufficiently simple to be effectively employed in both analytical and scalable simulative studies of caching systems. We demonstrate this by analytically characterizing the performance of the LRU caching policy under the SNM, for both a single cache and a network of caches. With respect to the standard independent reference model (IRM), some paradigmatic shifts, concerning the impact of various traffic characteristics on cache performance, clearly emerge from our results.

Cost effectiveness is critical to the success of immunisation programmes in developing countries. Being able to continuously, flexibly and accurately monitor the vaccine supply chain can help reduce the cost of operating it, and savings can be applied to deliver further resources to those that need

Content popularity prediction finds application in many areas, including media advertising, content caching, movie revenue estimation, traffic management and macro-economic trends forecasting, to name a few. However, predicting this popularity is difficult due to, among others, the effects of external phenomena, the influence of context such as locality and relevance to users,and the difficulty of forecasting information cascades.

No abstract available.

Distributed trust and reputation management mechanisms are often proposed as a means of providing assurance in dynamic and open environments by enabling principals to building up knowledge of the entities with which they interact. However, there is a tension between the preservation of privacy (which would suggest a refusal to release information) and the controlled release of information that is necessary both in order to accomplish tasks and to provide a foundation for the assessment of trustworthiness. However, if reputation-based systems are to be used in assessing the risks of privacy violation, it is necessary both to discover when sensitive information has been released, and then to be able to evaluate the likelihood that each of the set of principals that knew that information was involved in its release. We argue that statistical traceability can act as a basis for reaching a proper balance between privacy and trust. To enable this, we assume that interacting principals negotiate service level agreements that are intended to constrain the ways in which personal information may be used, and then monitor violations, ascribing likelihoods of involvement in release using an approach based on statistical disclosure control. Even though our approach cannot guarantee perfect privacy protection for personal information, it provides a framework using which detected privacy violation can be mapped onto a measure of accountability, which is useful in deterring such violation.