Brain tumour segmentation is a key application of AI in neuroimaging. Recently, federated learning (FL) has emerged as a strategic and increasingly relevant paradigm in neural computing due to its ability to address key challenges in large-scale neural network training, such as data access, privacy, collaborative learning, and model robustness. However, its adoption is currently hindered by high communication costs and the heterogeneity of client data. In this study, we investigated an efficient FL framework for brain tumour segmentation based on communication-aware optimization. We evaluated FedWSOComp, which integrates sparsification, quantization, and entropy-based encoding, in combination with a 3D U-Net architecture under both homogeneous and heterogeneous data distributions. The multi-institutional FeTS 2024 dataset was employed and partitioned into independent and identically distributed (IID) and non-IID settings, with an independent test set of 67 patients. An overall of 18 configurations combined sparsification rates and quantization levels. Performance was measured using Dice Similarity Coefficient (DSC) and 95th percentile Hausdorff Distance (HD95). Experimental results demonstrated that aggressive compression caused severe degradation in segmentation quality, with HD95 exceeding 60 mm. In contrast, higher retention with finer quantization achieved the best balance between efficiency and accuracy, reaching a DSC and HD95 mm on the test set under non-IID conditions. The findings demonstrated that, when configured with moderate-to-fine quantization and high sparsification retention, FedWSOComp enabled accurate and communication-efficient federated brain tumour segmentation. This study provides quantitative evidence and practical guidance for the deployment of FL-based segmentation models in privacy-sensitive and bandwidth-constrained clinical settings. • Analyse the impact of FedWSOComp, an integrated strategy combining top-k sparsification, quantization, and entropy-based encoding. • The performance of 18 different configurations (including IID and non-IID) was evaluated systematically. • High retention (60%) with fine quantization (64 clusters) optimizes performance.

Smart contracts are increasingly being adopted in modern supply chain (SC) systems, offering a transformative shift from traditional centralized models to decentralized, automated, and trustless processes. However, vulnerabilities in the chain of smart contract represent a critical threat to the reliability and security of supply chain ecosystems, often arising from intricate logical flaws, unintended inter-contract interactions, or improper handling of user input—issues that remain difficult to uncover through conventional testing or manual auditing. This paper introduces an innovative deep learning–based methodology that transforms smart contract bytecode into image representations, enabling precise and efficient classification of diverse vulnerability patterns. Unlike existing approaches that rely on source code availability or dynamic execution, the proposed framework operates independently of source code and circumvents the limitations inherent to dynamic analysis, thereby offering a versatile and system-agnostic solution. Experimental evaluations conducted on a newly curated dataset collected from multiple publicly available repositories demonstrate the robustness of the proposed method, achieving 92.24% accuracy and an 89.06% F1-score. Beyond its strong empirical performance, the framework ensures reproducibility, data transparency, and adaptability across heterogeneous blockchain environments. Collectively, these contributions establish a comprehensive and accessible foundation for enhancing the detection, mitigation, and overall resilience of blockchain smart contracts.

Federated Learning (FL) facilitates decentralized model training while prioritizing data privacy. However, its effective implementation faces significant challenges, primarily related to high communication overhead and energy consumption. Quantization, a key optimization technique, plays a crucial role in enhancing energy efficiency by reducing the bit precision of model updates, thereby lowering computational and transmission costs. This paper explores the impact of quantization on energy-efficient FL, focusing on techniques such as lattice quantization and stochastic gradient quantization methods. By compressing gradient updates and model parameters, quantization significantly reduces bandwidth requirements, enables efficient model aggregation, and prolongs battery life in resource-constrained edge devices. Furthermore, we discuss the trade-offs between quantization levels, model accuracy, and energy savings, emphasizing strategies to mitigate performance degradation while maintaining robust learning. Experimental results demonstrate that quantized FL can achieve up to 50% reduction in energy consumption while maintaining competitive model accuracy. This study highlights the importance of quantization in scalable, sustainable, and energy-aware FL, paving the way for its widespread adoption in real-world applications such as smart healthcare, IoT, and edge AI.

In today's interconnected cyber landscape, Distributed Denial of Service (DDoS) attacks represent a significant threat to the smooth functioning of online infrastructures. The nature of DDoS attacks, characterized by their distributed and dynamic nature, poses significant challenges for traditional centralized approaches to model training; however, the challenges of collaborative DDoS detection are compounded by stringent data privacy regulations, leaving mitigation efforts largely reliant on standalone and inflexible firewalls. Federated Learning (FL) represents a cutting-edge innovation in cybersecurity, presenting a revolutionary method for collectively training deep learning models without compromising sensitive data. Despite its promise, practical hurdles remain, particularly the reliance of most FL algorithms on centralized, server-side data for model evalu-ation-though some approaches avoid this centralized testing dependency. This limitation hinders the applicability of FL, especially in scenarios involving zero-day attacks on clients. Our paper examines a key hypothesis: whether the aggregated information from multiple clients can be effectively utilized to develop a global model that is inherently more resilient to zeroday attacks compared to models trained solely on individual client data. To investigate this, we introduce a methodology wherein FL models are trained on established DDoS attacks and subsequently evaluated against entirely novel, unencountered attacks, simulating zero-day scenarios at the client level. To ensure that each client contributes effectively to the training process, we utilize Jensen-Shannon Divergence (JSD) to evaluate and filter client updates based on their alignment with the global model. Building on this, we implement a kernel density estimation-based aggregation method to effectively mitigate feature distribution bias-a common issue in DDoS detection within FL environments. This approach forms a core component of our proposed framework, FLAME, which is built using the distributed framework Flower to realistically simulate FL in a decentralized setting. The code for our implementation can be found at: https://github.com/MODAL-UNINA/FLAME.

Human Activity Recognition (HAR) represents a significant domain within pervasive computing, facilitating a diverse array of applications ranging from healthcare to smart environments. Traditional HAR models suffer from several challenges, including data privacy and the distributed participation of heterogeneous resource-constrained devices. To mitigate these challenges, the research community popularly uses federated learning (FL). However, selecting clients in FL is a critical issue, mainly when there is a combination of resource-constrained heterogeneous devices. This paper proposes a resource-and performance-aware client selection algorithm for HAR, namely EAPD-CS, that amalgamates the benefits of FL with energy efficiency. The framework allows for the training of machine learning models across multiple devices without the necessity of sharing raw data, thereby preserving user privacy. Additionally, it employs energy-aware strategies to diminish the carbon footprint and reduce the computational costs typically linked to traditional cloud-based HAR systems. Experimental results indicate that the proposed framework achieves more than 90% accuracy, comparable to centralized models, while significantly lowering energy consumption and improving the robustness of the model. This work contributes to the evolving field of green AI, delivering an effective, privacy-preserving, and environmentally sustainable approach for HAR applications.

Analyzing big data using federated learning (FL) requires distributing the data to different clients to process locally and sending the model parameters to the global server for aggregation. In a real scenario, big data distribution is nonindependent and identically distributed (non-IID). Aggregating client updates is an important task in federated learning. Federated averaging (FedAvg) is the simplest and most popularly used method in FL. However, it is unable to handle heterogeneous (non-IID) data. To address this, in this paper we propose a Client Specific Dynamic Aggregation (CSDA) strategy focusing on dynamic aggregation based on client-specific metrics, aiming to improve the robustness and performance of the global model. Each client’s contribution is weighted by the quality and performance of their local updates, enhancing the overall federated learning process. Extensive experimentation has demonstrated that the proposed CSDA strategy, in conjunction with advanced comparison techniques, has the potential to greatly enhance the accuracy of each client across three real-world datasets.

Federated learning (FL) is a distributed machine learning (ML) approach that enables models to be trained on client devices while ensuring the privacy of user data. Model aggregation, also known as model fusion, plays a vital role in FL. It involves combining locally generated models from client devices into a single global model while maintaining user data privacy. However, the accuracy and reliability of the resulting global model depend on the aggregation method chosen, making the selection of an appropriate method crucial. Initially, the simple averaging of model weights was the most commonly used method. However, due to its limitations in handling low-quality or malicious models, alternative techniques have been explored. As FL gains popularity in various domains, it is crucial to have a comprehensive understanding of the available model aggregation techniques and their respective strengths and limitations. However, there is currently a significant gap in the literature when it comes to systematic and comprehensive reviews of these techniques. To address this gap, this paper presents a systematic literature review encompassing 201 studies on model aggregation in FL. The focus is on summarizing the proposed techniques and the ones currently applied for model fusion. This survey serves as a valuable resource for researchers to enhance and develop new aggregation techniques, as well as for practitioners to select the most appropriate method for their FL applications.

Numerous sensors on smart devices have made it possible to automatically recognize human movement, which might be helpful for intelligent applications like elder care, smart homes, and health monitoring. Nevertheless, implementing an activity recognition model in practical situations faces two main obstacles. First, machine learning models use a large number of labeled data to recognize human activities, which is not always feasible in real scenarios. Second, existing human activity recognition (HAR) systems cannot dynamically adapt to a new action. Furthermore, current methods fail to separate short-term activities from heterogeneous smart devices with varying positions and orientations that have similar sensory reading patterns. To address these issues, we propose Flexi-HAMR, an intelligent adaptive human activity monitoring and recognition system that dynamically recognizes activities using online, real-time activity signals. Many empirical findings show that the suggested flexible activity recognition model performs competitively on multiindividual activity identification tasks and has a comparatively more vital generalization ability.

The expeditious propagation of Internet of Things (IoT) technologies implanted in different smart devices such as smartphones and smartwatches have a ubiquitous consequence on the modern population. These devices are employed to collect data and to aid in tracking and analyzing the users’ daily activities using various human activity monitoring and recognition (HAR) techniques. However, most current HAMR approaches rely on exploratory case-based shallow feature learning architectures, which endeavor to recognize activities correctly in real-world situations. To address this issue, we offer a unique strategy for HAMR that leverages the attention mechanism with multi-head convolutional neural networks (CNNs) and Long-Short-Term-Memory (LSTM). The accuracy of activity detection is improved in the presented method by integrating attention into multi-head CNNs followed by LSTM for better feature extraction and selection. Verification investigations are carried out using data from the University of California (UCI) repository, which is publicly available. The results show that our proposed framework is more accurate than current frameworks using both the 10-fold and leave-one-subject-out cross-validation. Finally, the proposed method can recognize human activity in real-time, regardless of the type of smart device.

Declarative approaches to control-flow modeling use logic-based languages to formalize a number of constraints that valid traces must satisfy. The most noticeable example is the DECLARE framework based on linear temporal logic. Despite the interest that DECLARE has been attracting, the current knowledge about its formal properties was rather limited. The goal of this paper is to fill this gap by: (i) analyzing the behavioral properties of DECLARE by comparing it with the modeling capabilities of traditional procedural design approaches, in particular, block-structured processes; (ii) analyzing DECLARE from the computational point of view. As for the former point, we identify both the block-structured processes constructs that can be simulated in DECLARE and the features of DECLARE that can be encoded in block-structured processes. As for the latter point, we show that checking whether a given set of DECLARE patterns admits a satisfying trace is an NP-hard problem. In particular, we identify some DECLARE specifications whose satisfying traces are all of exponential length and some useful DECLARE fragments where a satisfying trace whose length is polynomially bounded is guaranteed to exist. The paper also discusses the declare2sat prototype system and the results of a thorough experimental validation.

Coalition structure generation is considered in a setting where feasible coalition structures must satisfy constraints of two different kinds modeled in terms of a valuation structure, which consists of a set of pivotal agents that are pairwise incompatible, plus an interaction graph prescribing that a coalition C can form only if the subgraph induced over the nodes/agents in C is connected. It is shown that valuation structures can be used to model a number of relevant problems in real-world applications. Moreover, complexity issues arising with them are studied, by focusing in particular on identifying islands of tractability based on topological properties of the underlying interaction graph. Stability issues on valuation structures are studied too.

We propose a hypergraph-based framework for modeling and detecting malevolent activities. The proposed model supports the specification of order-independent sets of action symbols along with temporal and cardinality constraints on the execution of actions. We study and characterize the problems of consistency checking, equivalence, and minimality of hypergraph-based models. In addition, we define and characterize the general activity detection problem, that amounts to finding all subsequences that represent a malevolent activity in a sequence of logged actions. Since the problem is intractable, we also develop an index data structure that allows the security expert to efficiently extract occurrences of activities of interest.

Process discovery has emerged as a powerful approach to support the analysis and the design of complex processes. It consists of analyzing a set of traces registering the sequence of tasks performed along several enactments of a transactional system, in order to build a process model that can explain all the episodes recorded over them. An approach to accomplish this task is presented that can benefit from the background knowledge that, in many cases, is available to the analysts taking care of the process (re-)design. The approach is based on encoding the information gathered from the log and the (possibly) given background knowledge in terms of precedence constraints , that is, of constraints over the topology of the resulting process models. Mining algorithms are eventually formulated in terms of reasoning problems over precedence constraints, and the computational complexity of such problems is thoroughly analyzed by tracing their tractability frontier. Solution algorithms are proposed and their properties analyzed. These algorithms have been implemented in a prototype system, and results of a thorough experimental activity are discussed.

Inverse frequent set mining (IFM) is the problem of computing a transaction database D satisfying given support constraints for some itemsets, which are typically the frequent ones. This article proposes a new formulation of IFM, called IFM I (IFM with infrequency constraints ), where the itemsets that are not listed as frequent are constrained to be infrequent; that is, they must have a support less than or equal to a specified unique threshold. An instance of IFM I can be seen as an instance of the original IFM by making explicit the infrequency constraints for the minimal infrequent itemsets, corresponding to the so-called negative generator border defined in the literature. The complexity increase from PSPACE (complexity of IFM) to NEXP (complexity of IFM I ) is caused by the cardinality of the negative generator border, which can be exponential in the original input size. Therefore, the article introduces a specific problem parameter κ that computes an upper bound to this cardinality using a hypergraph interpretation for which minimal infrequent itemsets correspond to minimal transversals. By fixing a constant k , the article formulates a k -bounded definition of the problem, called k -IFM I , that collects all instances for which the value of the parameter κ is less than or equal to k —its complexity is in PSPACE as for IFM. The bounded problem is encoded as an integer linear program with a large number of variables (actually exponential w.r.t. the number of constraints), which is thereafter approximated by relaxing integer constraints—the decision problem of solving the linear program is proven to be in NP. In order to solve the linear program, a column generation technique is used that is a variation of the simplex method designed to solve large-scale linear programs, in particular with a huge number of variables. The method at each step requires the solution of an auxiliary integer linear program, which is proven to be NP hard in this case and for which a greedy heuristic is presented. The resulting overall column generation solution algorithm enjoys very good scaling as evidenced by the intensive experimentation, thereby paving the way for its application in real-life scenarios.

The high-order coclustering problem, i.e., the problem of simultaneously clustering heterogeneous types of domain, has become an active research area in the last few years, due to the notable impact it has on several application scenarios. This problem is generally faced by optimizing a weighted combination of functions measuring the quality of coclustering over each pair of domains, where weights are chosen based on the supposed reliability/relevance of their correlation. However, little knowledge is likely to be available, in practice, in order to set these weights in a definite and precise manner. And, more importantly, it might even be conceptually unclear whether to prefer a weighing scheme over others, in those cases where functions encode contrasting goals so that improving the quality for a pair of domains leads to a deterioration for other pairs. The aim of this paper is precisely to shed light on the impact of weighting schemes on techniques based on linear combinations of pairwise objective functions, and to define an approach that overcomes the above problems by looking for an agreement-intuitively, a kind of compromise-among the various domains, thereby getting rid of the need to define an appropriate weighting scheme. Two algorithms performing coclustering on "star-structured” domains, based on linear combinations and agreements, respectively, have been designed within an information-theoretic framework. Results from a thorough experimentation, on both synthetic and real data, are discussed, in order to assess the effectiveness of the approaches and to get more insight into their actual behavior.

The inverse frequent set mining problem is the problem of computing a database on which a given collection of itemsets must emerge to be frequent. Earlier studies focused on investigating computational and approximability properties of this problem. In this paper, we face it under the pragmatic perspective of defining heuristic solution approaches that are effective and scalable in real scenarios. In particular, a general formulation of the problem is considered where minimum and maximum support constraints can be defined on each itemset, and where no bound is given beforehand on the size of the resulting output database. Within this setting, an algorithm is proposed that always satisfies the maximum support constraints, but which treats minimum support constraints as soft ones that are enforced as long as possible. A thorough experimentation evidences that minimum support constraints are hardly violated in practice, and that such negligible degradation in accuracy (which is unavoidable due to the theoretical intractability of the problem) is compensated by very good scaling performances.

Process-oriented systems have been increasingly attracting data mining researchers, mainly due to the advantages that the application of inductive process mining techniques to log data could open to both the analysis of complex processes and the design of new process models.

Process mining techniques have recently received notable attention in the literature; for their ability to assist in the (re)design of complex processes by automatically discovering models that explain the events registered in some log traces provided as input. Following this line of research, the paper investigates an extension of such basic approaches, where the identification of different variants for the process is explicitly accounted for, based on the clustering of log traces. Indeed, modeling each group of similar executions with a different schema allows us to single out "conformant" models, which, specifically, minimize the number of modeled enactments that are extraneous to the process semantics. Therefore, a novel process mining framework is introduced and some relevant computational issues are deeply studied. As finding an exact solution to such an enhanced process mining problem is proven to require high computational costs, in most practical cases, a greedy approach is devised. This is founded on an iterative, hierarchical, refinement of the process model, where, at each step, traces sharing similar behavior patterns are clustered together and equipped with a specialized schema. The algorithm guarantees that each refinement leads to an increasingly sound mDdel, thus attaining a monotonic search. Experimental results evidence the validity of the approach with respect to both effectiveness and scalability.

This paper investigates the problem of mining unconnected patterns in workflows and presents for its solution two algorithms, both adapting the Apriori approach to the graphical structure of workflows. The first one is a straightforward extension of the level-wise style of Apriori whereas the second one introduces sophisticated graphical analysis of the frequencies of workflow instances. The experiments show that graphical analysis improves the performance of pattern mining by dramatically pruning the search space of candidate patterns.

Today's workflow management systems represent a key technological infrastructure for advanced applications that is attracting a growing body of research, mainly focused in developing tools for workflow management, that allow users both to specify the "static" aspects, like preconditions, precedences among activities, and rules for exception handling, and to control its execution by scheduling the activities on the available resources. This paper deals with an aspect of workflows which has so far not received much attention even though it is crucial for the forthcoming scenarios of large scale applications on the Web: providing facilities for the human system administrator for identifying the choices performed more frequently in the past that had lead to a desired final configuration. In this context, we formalize the problem of discovering the most frequent patterns of executions, i.e., the workflow substructures that have been scheduled more frequently by the system. We attacked the problem by developing two data mining algorithms on the basis of an intuitive and original graph formalization of a workflow schema and its occurrences. The model is used both to prove some intractability results that strongly motivate the use of data mining techniques and to derive interesting structural properties for reducing the search space for frequent patterns. Indeed, the experiments we have carried out show that our algorithms outperform standard data mining algorithms adapted to discover frequent patterns of workflow executions.

This paper presents a rule-based declarative database language which extends DATALOG to express events and nondeterministic state transitions, by using the choice construct to model uncertainty in dynamic rules. The proposed language, called Event Choice DATALOG (DATALOG!ev for short), provides a powerful mechanism to formulate queries on the evolution of a knowledge base, given a sequence of events envisioned to occur in the future. A distinguished feature of this language is the use of multiple spatio-temporal dimensions in order to model a finer control of evolution. A comprehensive study of the computational complexity of answering DATALOG!ev queries is reported.