Alzheimer's disease (AD) is an irreversible neurodegenerative disorder and a leading cause of death worldwide. Early diagnosis plays an important part especially at the Mild Cognitive Impairment stage, where timely intervention can help slow its progression before it advances to AD. Neuroimaging data, like Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) scans, can help detect brain changes early by providing structural and functional brain changes related to the disease. Yet, many multimodal models still fuse MRI and PET with static concatenation and apply identical computation to all subjects, which limits robustness to patient/site heterogeneity and can waste computation. To address these limitations, we present the first study of combining 3D convolutional feature extractors with three fusion strategies - concatenation, Gated Multimodal Unit (GMU), and gated self-attention - and a sparsely gated Mixture-of-Experts (MoE) classifier that performs input-adaptive routing, activating only the most informative experts per case. Finally, we utilize Grad-CAM to visualize disease-related regions, ensuring model interpretability. Experiments are performed across three binary classification tasks (NC vs. MCI, MCI vs. AD, and NC vs. AD). Results show that GMU achieves accuracies of 80.46 % (NC vs. MCI) and 95.47 % (NC vs. AD), while gated self-attention attains 82.08 % on MCI vs. AD. Ablations show that removing the MoE consistently degrades accuracy across all tasks. These findings underscore the value of input-adaptive, multimodal modeling for AD diagnosis by leveraging the complementary nature of MRI and PET.

The rapid advancement of 5G networks is reshaping wireless communications through ultra-fast speeds, low latency, and seamless connectivity. This shift is accompanied by emerging technologies such as edge computing, network traffic management, resource allocation, and network slicing, which distribute data processing across the network. These trends amplify the need for privacy-preserving and decentralized learning methods, particularly in security-critical applications where transmitting raw data to a central server may be infeasible or undesirable. Federated Learning has emerged as a promising paradigm to meet these demands by enabling model training across distributed data sources. In this study, we explore the trustworthiness of Federated Learning models compared to a centralized counterpart, under various heterogeneous client data distributions. The distributions are generated via label-based symmetric Latent Dirichlet Allocations, where the Dirichlet concentration parameter controls the degree of class imbalance across clients. We use traffic flow data from a 5G Network Intrusion Detection task to design centralized and federated Artificial Neural Network architectures and extract feature importance scores using the Integrated Gradients algorithm. Our findings, based on the top-10 features, show that federated models trained on slightly-skewed ( ) and mildly-skewed ( ) data achieve trust scores closely aligned with the central model. The model achieves an average importance score of 2.2 (7.2% lower than the central model at 2.37), with 91.3% feature overlap. The model scores 2.44 (3% higher), with identical overlap. In contrast, highly-skewed models ( ) show diminished trustworthiness, scoring 1.62 and 1.74 (31.6% and 26.6% lower), with overlaps of 80% and 82.5%, respectively. These results highlight the impact of client data heterogeneity on model trustworthiness and underscore the sensitivity of federated models to high levels of data heterogeneity.

In today’s fast-paced world, the rates of stress and depression present a surge. People use social media for expressing their thoughts and feelings through posts. Therefore, social media provide assistance for the early detection of mental health conditions. Existing methods mainly introduce feature extraction approaches and train shallow machine learning (ML) classifiers. For addressing the need of creating a large feature set and obtaining better performance, other research studies use deep neural networks or language models based on transformers. Despite the fact that transformer-based models achieve noticeable improvements, they cannot often capture rich factual knowledge. Although there have been proposed a number of studies aiming to enhance the pretrained transformer-based models with extra information or additional modalities, no prior work has exploited these modifications for detecting stress and depression through social media. In addition, although the reliability of a machine learning (ML) model’s confidence in its predictions is critical for high-risk applications, there is no prior work taken into consideration the model calibration. To resolve the above issues, we present the first study in the task of depression and stress detection in social media, which injects extra-linguistic information in transformer-based models, namely, bidirectional encoder representations from transformers (BERT) and MentalBERT. Specifically, the proposed approach employs a multimodal adaptation gate for creating the combined embeddings, which are given as input to a BERT (or MentalBERT) model. For taking into account the model calibration, we apply label smoothing. We test our proposed approaches in three publicly available datasets and demonstrate that the integration of linguistic features into transformer-based models presents a surge in performance. Also, the usage of label smoothing contributes to both the improvement of the model’s performance and the calibration of the model. We finally perform a linguistic analysis of the posts and show differences in language between stressful and nonstressful texts, as well as depressive and nondepressive posts.

Although not all bots are malicious, the vast majority of them are responsible for spreading misinformation and manipulating the public opinion about several issues, i.e., elections and many more. Therefore, the early detection of bots is crucial. Although there have been proposed methods for detecting bots in social media, there are still substantial limitations. For instance, existing research initiatives still extract a large number of features and train traditional machine learning algorithms or use GloVe embeddings and train LSTMs. However, feature extraction is a tedious procedure demanding domain expertise. Also, language models based on transformers have been proved to be better than LSTMs. Other approaches create large graphs and train graph neural networks requiring in this way many hours for training and access to computational resources. To tackle these limitations, this is the first study employing only the user description field and images of three channels denoting the type and content of tweets posted by the users. Firstly, we create digital DNA sequences, transform them to 3d images, and apply pretrained models of the vision domain, including EfficientNet, AlexNet, VGG16, etc. Next, we propose a multimodal approach, where we use TwHIN-BERT for getting the textual representation of the user description field and employ VGG16 for acquiring the visual representation for the image modality. We propose three different fusion methods, namely concatenation, gated multimodal unit, and crossmodal attention, for fusing the different modalities and compare their performances. Finally, we present a qualitative analysis of the behavior of our best performing model. Extensive experiments conducted on the Cresci’17 and TwiBot-20 datasets demonstrate valuable advantages of our introduced approaches over state-of-the-art ones.

Alzheimer’s dementia (AD) affects memory, thinking, and language, deteriorating person’s life. An early diagnosis is very important as it enables the person to receive medical help and ensure quality of life. Therefore, leveraging spontaneous speech in conjunction with machine learning methods for recognizing AD patients has emerged into a hot topic. Most of the previous works employ Convolutional Neural Networks (CNNs), to process the input signal. However, finding a CNN architecture is a time-consuming process and requires domain expertise. Moreover, the researchers introduce early and late fusion approaches for fusing different modalities or concatenate the representations of the different modalities during training, thus the inter-modal interactions are not captured. To tackle these limitations, first we exploit a Neural Architecture Search (NAS) method to automatically find a high performing CNN architecture. Next, we exploit several fusion methods, including Multimodal Factorized Bi-linear Pooling and Tucker Decomposition, to combine both speech and text modalities. To the best of our knowledge, there is no prior work exploiting a NAS approach and these fusion methods in the task of dementia detection from spontaneous speech. We perform extensive experiments on the ADReSS Challenge dataset and show the effectiveness of our approach over state-of-the-art methods.

Epilepsy constitutes a chronic noncommunicable disease of the brain affecting approximately 50 million people around the world. Most of the existing research initiatives propose methods for detecting and predicting epilepsy, which rely on the extraction of handcrafted features and the train of traditional machine learning classifiers. In this paper, we present two new methods to distinguish healthy, interictal, and ictal cases without the time-consuming procedure of feature extraction. Firstly, we apply the short-time fourier transform (STFT) to the single-channel electroencephalogram (EEG) signals and construct an image consisting of three channels. This image is passed through pretrained models, including AlexNet, DenseNet201, EfficientNet, ResNet18, etc. Secondly, we introduce a multimodal deep neural network. Specifically, we pass each single-channel EEG signal through two branches of convolutional neural networks (CNNs), which can extract low and high frequency features. Also, we apply the short-time fourier transform (STFT) to the EEG signals and create an image consisting of three channels. The image is passed through a pretrained EfficientNet-B7 model. Finally, we employ a gated multimodal unit to control the importance of each modality. We evaluate the performance of the proposed model on five different cases on the EEG database of the University of Bonn and show that our introduced model achieves comparable performance to state-of-the-art approaches.

The Energy Sector is highly targeted by cyber threats because of its inherent value and profitability. Recent reported security incidents verify its key playing role in the entire economic and societal concurrent reality. This paper aims to assess the cyber-security culture status of European representatives in the entire electrical power supply chain during the coronavirus pandemic and the Ukrainian war. An evaluation campaign has been carefully designed and held from 3rd March 2022 to 18th March 2022. During that period, participants from different Electrical Power and Energy Systems (EPES) organizations participated in the campaign. Gathered results were analyzed and co-examined using different techniques revealing important findings regarding the cyber-security status and resilience of individuals and organizations in the European EPES sector.

Times of crisis have long been combined with an increase in cybercrime, exploiting the general instability; therefore, in such times, systems and infrastructures face greater exposure to vulnerabilities. On top of that, the COVID-19 crisis has increased our reliance on the internet, while working-from-home has been the daily reality for a large proportion of the population worldwide. Increased cyber-security awareness becomes a necessity for everyone, starting from a more knowledgeable audience; IT professionals, and software engineers. In this context, this paper aims to assess the cyber-security culture readiness of representatives studying or working within a European Polytechnique Academic Institution, during the COVID-19 crisis. Towards that end, a targeted evaluation campaign was launched for two weeks, from 28th February 2022 to 13th March 2022. The campaign consisted of four questionnaires of increased difficulty and a phishing quiz, all assessing the security culture of the participants against three dimensions; their security attitude, their competency, and their actual behavior. The campaign results have been thoroughly analyzed, and the findings were unforeseen in many cases, supporting the identification of security awareness weaknesses and assisting in drafting targeted, customized training programs.

The exponential growth of networking capabilities including the Internet of Things (IoT), has led to an outburst of cyberattacks. Many well-documented cyber-attacks have targeted critical energy infrastructures as well as any kind of cloud-based IT platforms. Early examination of critical systems’ vulnerabilities, as well as previous cyber-security incidents, are of utmost importance to prevent new ones. A thorough investigation to examine the context of the cyber-security breach can reveal facts about the source of the attack, the profile of the attacker, the resources, and the skills required and can further reveal mitigations for preventing the attack from re-appearing in the future. To safeguard critical energy infrastructures, many forensic approaches have been developed to collect, analyze, and digitalize evidence assisting in the in-depth investigation of an incident. However, up to now, the many open-source vulnerability data sources which have been developed to provide valuable information for a cyber-attack are yet to be employed to assist in forensic investigation. This paper introduces the Automated Forensic Tool, a platform that employs machine learning algorithms to combine different vulnerability data sources for facilitating the forensic procedure while minimizing the time and effort needed. A use case is also demonstrated that displays how the tool can be used towards assisting the forensic investigation of cyber-security incidents on an energy infrastructure, but the tool can also be applied to other critical energy and IT infrastructures with minor adaptations.

Insider threat has been recognized by both scientific community and security professionals as one of the gravest security hazards for private companies, institutions, and governmental organizations. Extended research on the types, associated internal and external factors, detection approaches and mitigation strategies has been conducted over the last decades. Various frameworks have been introduced in an attempt to understand and reflect the danger posed by this threat, whereas multiple identified cases have been classified in private or public databases. This paper aims to present how a cyber-security culture framework with a clear focus on the human factor can assist in detecting possible threats of both malicious and unintentional insiders. We link current insider threat categories with specific security domains of the framework and introduce an assessment methodology of the core contributing parameters. Specific approach takes into consideration technical, behavioral, cultural, and personal indicators and assists in identifying possible security perils deriving from privileged individuals.

This paper presents a cyber-security culture framework for assessing and evaluating the current security readiness of an organization’s workforce. Having conducted a thorough review of the most commonly used security frameworks, we identify core security human-related elements and classify them by constructing a domain agnostic security model. We then proceed by presenting in detail each component of our model and attempt to quantify them in order to achieve a feasible assessment methodology. The paper thereafter presents the application of this methodology for the design and development of a security culture evaluation tool, that offers recommendations and alternative approaches to workforce training programs and techniques. The model has been designed to easily adapt on various application domains while focusing on their unique characteristics. The paper concludes on applications of our instrument on security-critical domains, and its contribution to current research by providing deeper insights regarding the human factor in cybersecurity.

Alzheimer's disease (AD) is the main cause of dementia which is accompanied by loss of memory and may lead to severe consequences in peoples' everyday life if not diagnosed on time. Very few works have exploited transformer-based networks and despite the high accuracy achieved, little work has been done in terms of model interpretability. In addition, although Mini-Mental State Exam (MMSE) scores are inextricably linked with the identification of dementia, research works face the task of dementia identification and the task of the prediction of MMSE scores as two separate tasks. In order to address these limitations, we employ several transformer-based models, with BERT achieving the highest accuracy accounting for 87.50%. Concurrently, we propose an interpretable method to detect AD patients based on siamese networks reaching accuracy up to 83.75%. Next, we introduce two multi-task learning models, where the main task refers to the identification of dementia (binary classification), while the auxiliary one corresponds to the identification of the severity of dementia (multiclass classification). Our model obtains accuracy equal to 86.25% on the detection of AD patients in the multi-task learning setting. Finally, we present some new methods to identify the linguistic patterns used by AD patients and non-AD ones, including text statistics, vocabulary uniqueness, word usage, correlations via a detailed linguistic analysis, and explainability techniques (LIME). Findings indicate significant differences in language between AD and non-AD patients.

In order to get a deeper understanding of security breaches, their severity, impact and ways to mitigate them, many vulnerability databases and dictionaries have been developed. However, all that information on vulnerabilities is scattered all over the web, which makes locating and mitigating vulnerabilities an arduous task. This paper introduces the Vulnerabilities Manager, a tool that automates the process of linking information from well-known external vulnerability data sources. Its goal is to present an enriched vulnerability report to its final users, assisting them in pinpointing software and hardware assets’ defects, categorizing and prioritizing them, thus, contributing to the cyber defense against potential security breaches and adversary actions. To achieve this, the Vulnerabilities Manager exploits current state of the art machine learning and artificial intelligence techniques. The tool may also be enriched with forensic capabilities, detecting cyber threats, unveiling information about the nature of the attacker, and proposing mitigations against them in real-time.

This paper deals with the process of crowdsourcing requirements elicitation in software engineering and the alignment of the customer needs during the development phase, through the usage of anonymous personas, and the support of the persona builder application that allows the extraction of such information through anonymized data. Having identified the realization of users and customers' needs in the software engineering cycle, despite the adoption of agile methods, the paper suggests the usage of a persona that represents a set of users with similar characteristics, a pool of personas that software teams may share with each other through a collaborative application, and persona builder as a tool to generate such personas through real user profiles and data collected through third party services. At the end, a demo application is presented, realizing the concept of anonymized, crowdsourced personas.

Although customers become more and more vocal in expressing their experiences, demands and needs in various social networks, companies of any size typically fail to effectively gain insights from such social data and to eventually catch the market realm. This paper introduces the Anlzer analytics engine that aims at leveraging the "social" data deluge to help companies in their quest for deeper understanding of their products' perceptions as well as of the emerging trends in order to early embed them into their product design phase. The proposed approach brings together polarity detection and trend analysis techniques as presented in the architecture and demonstrated through a simple walkthrough in the Anlzer solution. The Anlzer implementation is by design domain-independent and is being tested in the furniture domain at the moment, yet it brings significant added value to software design and development, as well, through its experimentation playground that may provide indirect feedback on future software features while monitoring the reactions to existing releases.

The ever-accelerating growth of cloud-based services (CBS) and the prevalence of multi-sided business models have distributed users' data across different data silos that hinder mobile applications development and sustainability. The present paper aims at describing an open framework that abstracts functionality from CBSs through a common Graph, RESTful API, which manages calls among various CBS APIs and syndicates responses under a common standardized format. Combining this conceptual framework with semantically enriched modeling, the implemented platform allows a community of developers to govern, extend and maintain the Graph API and consequently, applications to access a plethora of CBSs through a single point of access. Building on the experience of third-party solutions that mash-up data from different services in their API, the proposed approach goes beyond the state-of-the art through its community-orientation, the API extensibility-by-design and the advanced context awareness and sophistication it provides to developers.

Today, a new age of engagement and collaboration has emerged with the proliferation of usergenerated content in social networks and generally the Web 2.0, rendering it particularly difficult for enterprises to monitor and act upon all content following conventional data mining methodologies. In this paper, we present our approach for a Future Internet enabler (FITMAN Anlzer) that provides automated, social data analytics and aims at assisting enterprises in becoming more tuned to their customer needs and gaining insights into current and future trends to early embed them into product design. The FITMAN Anlzer implementation is domainindependent and allows any manufacturer to effectively train it based on his needs and create personalized reports to timely capture the right information. Our methodology includes trend analytics, polarity detection through machine learning, data querying through flexible reports and finally informative charts to visualize the results in order to help companies in their decision making procedures.

Considering the fact that contemporary business settings call for work in groups, team selection and formation is a crucial parameter for the smooth function and therefore the achievement of the specific team and business objectives. The problem of team selection members is particularly complex due to the variety of factors and criteria that need to be taken into consideration. Thus, highlighting the complexity in selecting team members, this paper proposes a multi criteria approach to deal with group decision making under fuzzy environment. A Multi Criteria Decision Making Approach (the fuzzy TOPSIS) for group decision making is considered, incorporating a new measurement, which reflects the minimum requirements of the decision makers for each criterion. In this respect, a new reference point is introduced, apart from the Positive Ideal Solution and the Negative Ideal Solution. Finally, an illustrative example of the proposed approach is presented for the selection of a middle level consulting manager.

Purpose - This paper aims to deal with the development of a risk management application for the modelling, optimal adaptation and implementation of an ERP system. Design/methodology/approach - This paper presented a risk management application for the modeling, optimal adaptation and implementation of an ERP system. The application was tested with the operations and capabilities of the ERP commercial package "SINGULAR Enterprise (SEn)" of the Greek Software House DELTA-SINGULAR S.A. Findings - The functional result of this application was proved to support considerably the management of risk within the implementation of the ERP system. Originality/value - To the best knowledge of the authors there is no other current generic research in this technological field concerning small or medium-sized enterprises. With the development of this application, the goals mentioned in the conclusions were achieved. © Emerald Group Publishing Limited.

In this paper we consider the use of ontologies as the basis for structuring and simplifying the process of constructing real-time problem-solving tools, focusing specifically on the task of production scheduling. In spite of the commonality in production scheduling system requirements and design, different scheduling environments invariably present different challenges (e.g. different constraints, different objectives, different domain structure, etc.). The proposed methodology for building ontologies used for production scheduling systems represents a synthesis of extensive work in developing constraint-based scheduling models for a wide range of applications in manufacturing and production planning. Since the effective modeling is one of the most important and difficult steps in the development of reliable information systems, and taking into consideration the fact that the general problem of the production scheduling in the industries is very difficult and still unsolved, one can easily estimate the merit of this methodology.