Evasive poisoning attacks such as semantic backdoors pose a growing threat to federated learning because they mimic benign client updates and evade detectors under secure aggregation. We introduce an unsupervised, per-client structural check that runs after each local round, before aggregation, requiring only compact statistical summaries derived from each client update after a geometric transformation that removes dependence on the global model. Client-update statistics are compared against a calibrated benign reference, and deviations are detected through statistical distances. Energy and Wasserstein-1 jointly define an operational pattern where threshold exceedances across projections reveal structural deviations even in apparently benign updates. Evaluated on canonical backdoor scenarios from Bagdasaryan et al., the method detects both strong and stealthy attacks in the first local round through consistent multi-projection threshold excesses, while benign updates show only isolated ones. The procedure is lightweight, unsupervised, compatible with secure aggregation, and does not require trigger datasets. By providing early per-client warnings before aggregation, it complements classical defenses such as norm clipping, differential privacy, and robust aggregation, enabling proactive mitigation of poisoning in federated learning.

Organizations increasingly rely on digital platforms to support their operations, decision-making processes, and the delivery of critical services. This growing dependence has expanded their exposure to cyber threats that jeopardize the confidentiality, integrity, avail-ability, and operational continuity of information systems. This paper presents a system specifically designed to support the identi- fication and management of risks in technological infrastructures. The proposed solution collects vulnerability data daily from official sources and correlates it with the organization’s assets, enabling the prioritization of risks according to their criticality level. Further-more, the system integrates a prediction module based on machine learning techniques, capable of estimating the aggregated evolu- tion of risk for the following month. This predictive capability facilitates preventive decision-making and strengthens proactive cybersecurity risk management strategies.

Different modeling approaches have been proposed to overcome every design pitfall of the development of the different parts of a data warehouse (DW) system. However, they are all partial solutions which deal with isolated aspects of the DW and do not provide designers with an integrated and standard method for designing the whole DW (ETL processes, data sources, DW repository and so on). On the other hand, the Model Driven Architecture (MDA) is a standard framework for software development that addresses the complete life cycle of designing, deploying, integrating, and managing applications by using models in software development. In this paper, we describe how to align the whole DW development process to MDA. Then, we define MD2A (MultiDimensional Model Driven Architecture), an approach for applying the MDA framework to one of the stages of the DW development: multidimensional (MD) modeling. First, we describe how to build the different MDA artifacts (i.e. models) by using extensions of the Unified Modeling Language (UML). Secondly, transformations between models are clearly and formally established by using the Query/View/Transformation (QVT) approach. Finally, an example is provided to better show how to apply MDA and its transformations to the MD modeling.