Abstract
Modern communications have moved away from point-to-point models to increasingly heterogeneous network models. In this article, we propose a novel controller-based architecture to deploy adaptive causal network coding in heterogeneous and highly meshed communication networks. Specifically, we consider using the software-defined network as the main controller. We first present an architecture for highly meshed heterogeneous multi-source multi-destination networks that represent the practical communication networks encountered in the fifth generation of wireless networks and beyond. Next, we present a promising solution to deploy network coding over the new architecture. We also present a new controller-based setting with which network coding modules communicate to attain the required information. Finally, we briefly discuss how the proposed architecture and network coding solution provide a good opportunity for future technologies.

Abstract
Software Defined Networking (SDN) facilitates the orchestration and configuration of network resources in a flexible and scalable form, where policies are managed by controller components that interact with network elements through multiple interfaces. The ubiquitous adoption of SDN leads to the availability of multiple SDN controllers, which have different characteristics in terms of performance and security support. SDN controllers are a common target in network attacks since their compromise leads to the capability of impairing the entire network. Thus, the choice of a SDN controller must be a meticulous process from early phases (design to production). CROCUS, herein proposed, provides a mechanism to enable an objective assessment of the security support of SDN controllers. CROCUS relies on the information provided by the Common Vulnerability Scoring System (CVSS) and considers security features derived from scenarios with stringent security requirements. Considering a vehicular communication scenario supported by multiple technologies, we narrow the selection of SDN controllers to OpenDayLight and ONOS choices. The results put in evidence that both controllers have security features relevant for demanding scenarios with ONOS excelling in some aspects.

Abstract
Current methods for optimization of low-density parity-check (LDPC) codes analyze the degree distribution pair asymptotically as block length approaches infinity. This effectively ignores the discrete nature of the space of valid degree distribution pairs for LDPC codes of finite block length. While large codes are likely to conform reasonably well to the infinite block length analysis, shorter codes have no such guarantee. We present and analyze an algorithm for completely enumerating the space of all valid degree distribution pairs for a given block length, code rate, maximum variable node degree, and maximum check node degree. We then demonstrate this algorithm on an example LDPC code of finite block length. Finally, we discuss how the result of this algorithm can be utilized by discrete optimization routines to form novel methods for the optimization of small block length LDPC codes.

Abstract
Due to the pervasiveness of Interconnected devices, large amounts of heterogeneous data types are being continuously collected. Regardless of the benefits that come from sharing data, exposing sensitive and private information arises serious privacy concerns. To prevent unwanted disclosures and, hence, to protect users' privacy, several privacy-preserving mechanisms have been proposed. However, the data heterogeneity and the inherent correlations among the different data types have been disregarded when developing such mechanisms. Our goal is to develop privacy-preserving mechanisms that are suitable for data heterogeneity and data correlation. These aspects will also be considered to develop mechanisms to achieve private learning. © 2021 Owner/Author.

Abstract
Grade prediction has been for a long time a subject that interests both teachers and researchers. Before the digital age this type of predictions was something nearly impossible to achieve. With the increasing integration of Learning Management Systems in education, grade prediction seems to have become a viable option. The general adoption of this type of systems brings to the research area a database known as “registry”, or more simply known as logged data. Using this new source of information several attempts regarding the prediction of student grades have been proposed. The methodology proposed in this study is capable of, analyzing student online behavior, using the information collected by the Moodle system and making a prediction on what the final grade of the student will be, at any point in the semester. Our novel approach uses the gathered information to examine the academic path of the student in order to determine an interaction pattern, then it tries to establish a link with other, present or past, known successful paths. Making this comparison, the model can automatically determine if a student is going to fail or pass the course, which then would leave a space for the teacher or the student to circumvent the situation. Our results show that the system is not only viable, as it is also robust to make prediction at an early stage in the course.

Abstract
Social networks have provided the means for constant connectivity and fast information dissemination. In addition, real-time posting allows a new form of citizen journalism, where users can report events from a witness perspective. Therefore, information propagates through the network at a faster pace than traditional media reports it. However, relevant information is a small percentage of all the content shared. Our goal is to develop and evaluate models that can automatically detect journalistic relevance. To do it, we need solid and reliable ground truth data with a significantly large quantity of annotated posts, so that the models can learn to detect relevance over all the spectrum. In this article, we present and confront two different methodologies: an automatic and a human approach. Results on a test data set labelled by experts' show that the models trained with automatic methodology tend to perform better in contrast to the ones trained using human annotated data.