Modern systems face increasingly sophisticated threats, driving the need for detection frameworks capable of identifying complex behavioral anomalies. Through integrating neural architectures with flowgraph analysis, a versatile framework was developed to classify malicious activities with high accuracy and adaptability. Temporal dependencies and structural patterns in system behavior are dynamically analyzed to differentiate between benign and malicious operations, even in the presence of obfuscation and polymorphic strategies. Experimental results demonstrated consistent outperformance over traditional methods, particularly in terms of reducing false positive rates while maintaining robust classification precision. The modular design facilitates deployment across a variety of operational environments, from distributed systems to real-time detection scenarios. Scalability testing revealed that detection performance remained stable even with increasing data complexity, affirming its suitability for large-scale applications. Adversarial testing highlighted the framework's resilience against perturbation attacks, ensuring reliable performance under evolving threat landscapes. Efficiency metrics, including energy consumption and latency, further validated its practicality for enterpriselevel integration. Adaptability to diverse ransomware families, such as LockBit, Hive, and Conti, was demonstrated through thorough experiments, illustrating its versatility in addressing different attack vectors. Challenges posed by imbalanced datasets were effectively mitigated, showcasing the system's capacity to maintain accuracy in skewed conditions. Real-world applications extend to securing critical infrastructure, IoT ecosystems, and cloud platforms, where rapid and accurate threat detection is crucial. Neural-Mimetic Flowgraph Analysis emerges as a substantial advancement in cybersecurity, offering a robust solution to evolving digital threats.

The integration of artificial intelligence (AI) in healthcare, powered by Internet of Medical Things (IoMT) data, offers significant potential for personalized and efficient patient care. Hierarchical federated learning (HFL) is a promising approach for healthcare applications, combining cross-silo and cross-device federated learning. This architecture allows hospitals to train local models using patient data, while sharing anonymized parameters with other hospitals to improve diagnosis. However, existing studies on incentive mechanisms in HFL often focus on determining optimal incentive values but neglect the integration of these incentives into the reward stage. Moreover, the two-layer architecture of HFL introduces challenges related to disparities in patient volume and diversity across hospitals. In this paper, we propose a fair incentive distribution mechanism for hierarchical systems using blockchain state channels. We ensure equal incentive budget contributions from all organizations, preventing free-riders in the HFL system with a channel factory solution. Additionally, virtual channels support transactions without intermediaries, minimizing computational costs in the blockchain network.

Wireless sensing has made significant progress in tasks ranging from action recognition, vital sign estimation, pose estimation, etc. After over a decade of work, wireless sensing currently stands at the tipping point transitioning from proofof-concept systems to the large-scale deployment. We envision a future service scenario where wireless sensing service providers distribute sensing models to users. During usage, users might request new sensing capabilities. For example, if someone is away from home on a business trip or vacation for an extended period, they may want a new sensing capability that can detect falls in elderly parents or grandparents and promptly alert them. In this paper, we propose CCS (continuous customized service), enabling model updates on users' local computing resources without data transmission to the service providers. To address the issue of catastrophic forgetting in model updates-where updating model parameters to implement new capabilities leads to the loss of existing capabilities-we design knowledge distillation and weight alignment modules. These modules enable the sensing model to acquire new capabilities while retaining the existing ones. We conducted extensive experiments on the large-scale XRF55 dataset across Wi-Fi, millimeter-wave radar, and RFID modalities to simulate scenarios where four users sequentially introduced new customized demands. The results affirm that CCS excels in continuous model services across all the above wireless modalities, significantly outperforming existing approaches like OneFi [1].

In microwave remote sensing of vegetated land surfaces, there are volume scattering from the vegetation and rough-surface scattering of the interface between the vegetation and the ground. However, the coupling of volume scattering and surface scattering and its impact on microwave emission have not been fully studied, particularly when multiple scattering effects are to be considered as in forested scenarios. Earlier treatment with the Kirchhoff approximation neglects the incoherent bistatic scattering components from the rough soil interface and thus overestimates the transmitted upward emission from the soil half space and underestimate the reflected upward emission from the downward-going intensity streams arising from the vegetation layer in non-specular directions. To address this issue, bistatic scattering coefficients are introduced into the boundary conditions and an interpolation-accelerated numerical iterative method is employed to solve the passive radiative transfer equation to rigorously couple the volumesurface scattering interactions at the vegetation/soil interface. In this paper, the rough-surface bistatic scattering is modeled using the advanced integral equation method (AIEM). The volume scattering model is derived from a radiative transfer based multiple scattering model that accounts for the vertical profile of the vegetation structure. The SMAPVEX12 forest dataset is utilized to validate the proposed model, encompassing L-band radiometric brightness temperature observations corresponding to extensive variations in soil moisture. Comparisons are made between the brightness temperatures simulated by the model with and without considering the incoherent bistatic scattering coefficients, under diverse vegetation water contents (VWCs) and varying soil roughness conditions. Notable features are observed in the angular pattern of the vegetated surface emission by fully accounting for the multiple scattering effects and the incoherent bistatic scattering effects of the rough soil. These features become more evident at smaller observation angles, lower VWCs, and greater soil roughness. The findings reveal that the model incorporating the bistatic scattering coefficient of rough surfaces exhibits improved agreement with the measured brightness temperatures. Furthermore, the proposed model is parameterized by matching the high-order solutions to the RT equation to the widely adopted albedo-tau formalism, i.e., the zero-order solution to the passive radiative transfer equation with a flat lower boundary. The resulting equivalent optical thickness and the equivalent scattering albedo incorporates the multiple scattering effects within the vegetation layer and they are solely associated with the geometries and the electromagnetic properties of the vegetation layer. Additionally, the equivalent reflectivity of the soil is proposed to characterize the scattering properties of rough surfaces and the scattering coupling between the rough surface and the vegetation layer. The new model developments will significantly enhance the prediction and interpretation of vegetated land surface emission characteristics and thus improve the remote sensing of the vegetation and the underlying soil parameters.

The building and construction industry is a significant contributor to global greenhouse gas emissions, with 37% of total emissions attributed to this sector. While efforts have primarily targeted operational carbon emissions, which stem from building operations like heating and cooling, the urgent need to address embodied carbon—associated with the materials used and their life cycle—has gained attention. This paper explores the critical role of material selection and innovative practices in reducing embedded carbon in the built environment. It highlights collaborative models and international cooperation essential for decarbonizing building materials to achieve net-zero emissions by mid-century. The findings underscore that embodied carbon currently represents a growing proportion of a building’s overall carbon footprint, necessitating proactive measures in the design and construction phases. By integrating life cycle assessments and prioritizing sustainable material choices, stakeholders can significantly diminish carbon emissions and align with global climate goals. Through case studies and best practices, this research advocates for a comprehensive approach to carbon reduction that encompasses both operational and embodied emissions in the built environment.

Sentiment analysis is an important tool for revealing insights and shaping our understanding of market movements from financial articles, news, and social media. Despite their impressive abilities in financial natural language processing (NLP), large language models (LLMs) still have difficulties in accurately interpreting numerical values and grasping financial context, limiting their effectiveness in predicting financial sentiment. This article introduces a simple and effective instruction-tuning method to solve these problems. We have made significant progress in financial sentiment analysis by converting small amounts of supervised financial sentiment analysis data into command data and using this approach to fine-tune a generic LLM. In experiments, our approach outperforms state-of-the-art supervised sentiment analysis models and widely used LLMs such as ChatGPT and LLaMAs, especially when numerical and contextual understanding is critical.

Quantitative capillary-free zone (CFZ) analysis in optical coherence tomography angiography (OCTA) holds promise for the early detection of retinal diseases. However, its clinical deployment is hindered by time-consuming manual segmentation procedures for periarterial and perivenous CFZs, which often focus solely on large vessels. In this study, we introduce deep learning for the automated segmentation of periarterial and perivenous CFZs in OCTA. Both convolutional neural networks (CNNs) and vision transformers (ViTs) were evaluated for automated CFZ segmentation. Nine quantitative features were derived to characterize CFZ changes associated with diabetic retinopathy (DR). Quantitative analysis revealed significant changes in CFZ ratios, counts, and mean sizes that can reliably differentiate control subjects, diabetic patients without DR (NoDR), and those with mild DR, underscoring their potential as sensitive biomarkers for early disease detection, progression monitoring, and treatment assessment.

Recent advancements in artificial intelligence have expanded the possibilities for processing complex multi-modal documents, yet maintaining context coherence and factual accuracy across different modalities remains a significant challenge (Tu et al. 2024; Es et al. 2024; Klein et al. 2006a). Building upon foundational work in graph-based retrieval systems, particularly Microsoft's GraphRAG (Edge et al. 2024) and the open-source LightRAG (Guo et al. 2024) project, we present LuminiRAG, a system that advances document understanding through several key innovations in content extraction and query processing. Our primary contributions include: (1) an intelligent content relevance assessment mechanism that selectively extracts meaningful information from multi-modal documents (Ram et al. 2023; Fan et al. 2024), (2) an enhanced vision-language processing pipeline that preserves semantic relationships between textual and visual elements (Chan et al. 2024), (3) a dynamic knowledge graph construction approach that extends GraphRAG principles to maintain cross-modal relationships (Petroni et al. 2021), and (4) a sophisticated query processing engine that handles complex, multi-hop queries through adaptive semantic thresholding and comprehensive response verification (Gao et al. 2023). Through extensive evaluation on diverse document sets in the financial domain, including financial reports, transaction documents, and contractual documents, LuminiRAG demonstrates substantial improvements over traditional RAG approaches (Es et al. 2024; Salemi and Zamani 2024). Our results show particularly significant improvements in handling complex queries involving tabular data, cross-references, and multi-modal content relationships, establishing new benchmarks in document understanding while maintaining practical processing efficiency.

The emergence of distinct machine learning explanation methods has leveraged a number of new issues to be investigated. The disagreement problem is one such issue, as there may be scenarios where the output of different explanation methods disagree with each other. Although understanding how often, when, and where explanation methods agree or disagree is important to increase confidence in the explanations, few works have been dedicated to investigating such a problem. In this work, we proposed Visagreement, a visualization tool designed to assist practitioners in investigating the disagreement problem. Visagreement builds upon metrics to quantitatively compare and evaluate explanations, enabling visual resources to uncover where and why methods mostly agree or disagree. The tool is tailored for tabular data with binary classification and focuses on local feature importance methods. In the provided use cases, Visagreement turned out to be effective in revealing, among other phenomena, how disagreements relate to the quality of the explanations and machine learning model accuracy, thus assisting users in deciding where and when to trust explanations. To assess the effectiveness and practical utility of Visagreement, we conducted an evaluation involving four experts. These experts assessed the tool's Effectiveness, Usability, and Impact on Decision-Making. The experts confirm the Visagreement tool's effectiveness and user-friendliness, making it a valuable asset for analyzing and exploring (dis)agreements.

Fleets of autonomous vehicles offer an innovative solution to enhance logistic operations in manufacturing systems by increasing efficiency and safety. However, navigating in unstructured environments with moving obstacles presents significant challenges. This paper addresses this problem by proposing a novel control architecture that combines grid based mapping with receding horizon control techniques. Specifically, unicycle models are used to describe the dynamics of the robots, and sequences of feasible and safe way-points are computed by leveraging the real-time grid map of the entire operating scenario. A distributed set-theoretic model predictive control scheme leverages this information to ensure safe navigation. The effectiveness of the proposed approach is demonstrated through experimental results in a real-world context, where a team of three autonomous robots assists human workers along a steel production line manufacturing angle pipes, block-stages, and caps.

In industrial DC-DC power conversion applications, achieving high efficiency and small size in converters is crucial. To address these issues, this article introduces a novel semiinterleaved non-isolated topology by employing coupled inductors and a Valley-Fill structure. This method ensures continuous output current with zero-ripple, which reduces the size of the output filter. Additionally, by implementing soft switching techniques, the converter can operate at higher frequencies and accommodate a wide range of load variations. The utilization of a pulse frequency modulation (PFM) controller sustains zero-voltage switching (ZVS) and improves the conversion ratio by changing frequency. Also, self-driven synchronous rectifiers (SRs) turn ON and OFF under ZVS circumstances, and Valley-Fill diodes turn OFF under zero-current switching (ZCS) conditions, thereby lowering switching, diode, and reverse recovery losses, ultimately enhancing converter efficiency. A 300-V to 24-V/12-A prototype of the proposed converter is implemented to verify the performance and the theoretical analysis which results in efficiency higher than 97% from full load to light load.

Hierarchical contextual embedding is a novel approach to encoding semantic relationships within complex linguistic data, emphasizing the integration of token interactions across multiple levels of abstraction. By leveraging a multi-layered structure, the proposed framework successfully captures both local dependencies and global contextual patterns, addressing challenges associated with long-range dependencies and extended textual inputs. Extensive experiments reveal substantial performance improvements in tasks requiring deep contextual comprehension, supported through robust scalability and efficient handling of large input sequences. The embedding mechanism demonstrates adaptability across diverse linguistic domains, including tasks with zero-shot transfer learning scenarios, while maintaining competitive accuracy. Comparative analyses highlight the framework's advantages over baseline models, particularly in representing syntactic hierarchies and semantic relationships in a computationally efficient manner. Visualization of embedding spaces illustrates coherent clustering of related tokens, affirming the framework's ability to encode hierarchical linguistic structures effectively. Error analysis uncovers specific challenges, such as resolving nested clauses and ambiguous references, which demonstrate the complexity of achieving comprehensive linguistic modeling. Scalability tests validate the model's capacity to manage varying input sizes while maintaining computational efficiency. Results demonstrate the framework's robustness against noisy inputs, with consistent accuracy across moderate error rates. Theoretical principles underpinning the hierarchical approach align closely with observed outcomes, reinforcing the framework’s conceptual soundness. With its innovative design, the framework contributes significantly to the advancement of adaptive language modeling architectures and their practical applications. Comprehensive evaluations across tasks and domains affirm the framework’s broad applicability and potential to inspire future research in language modeling and contextual representation.

The rapid evolution of technology has brought forth unprecedented opportunities and challenges in this digital era. Ethical issues in responsible data research have become a principal concern among these challenges necessitating thoughtful consideration and proactive management (Esmer & Arıbaş, 2022). As data becomes a pivotal asset for organizations and individuals alike, the methodologies and practices encompassing its collection, analysis, and storage must be scrutinized to ensure ethical compliance (Edquist et al., 2022). The importance of ethics in responsible data research has been amplified within the realm of cybersecurity where researchers are tasked with securing sensitive information and protecting the privacy of individuals and organizations. Cybersecurity is not just about defending against malicious attacks but also about maintaining the trust of stakeholders through ethical conduct and moral lapses in cybersecurity can result in ramifications such as breaches of confidentiality, loss of data integrity, and violation of public trust (Macnish & Van der Ham, 2020). This implies providing a comprehensive understanding of the ethical deliberations cybersecurity researchers must navigate to balance security and moral standards in an effective manner. The examination of case studies, regulatory frameworks, and best practices can equip cybersecurity practitioners with the knowledge and tools required to uphold ethical standards in their work alongside underscoring the need for continuous ethical vigilance in a rapidly evolving digital landscape while ensuring that the data protection standards align with the broader societal values of fairness, privacy, and respect for individuals' rights (Miller & Bossomaier, 2024).

Decision-Making is a fundamental topic in the domain of Autonomous Driving where significant challenges must be tackled due to the variable behaviours of surrounding agents and the wide array of encountered scenarios. The primary aim of this work is to develop a hybrid Decision-Making architecture able to be validated on a real vehicle that marries the reliability of classical techniques with the adaptability of Deep Reinforcement Learning approaches. To address the crucial transition from simulated training environments to real-world applications, this research employs a Curriculum Learning approach, facilitated by the deployment of Digital Twins and Parallel Intelligence technologies, significantly narrowing the Reality Gap and enhancing the applicability of learned behaviors. The viability of this approach is evidenced through Parallel Execution, wherein simulated and real-world tests are conducted simultaneously. Specifically, our approach consistently exceeds the performance benchmarks established by existing frameworks in the Reinforcement Learning literature within SMARTS, achieving success rates greater than 91%. Furthermore, it completes various scenarios in CARLA up to 50% faster than Autopilot, demonstrating improved comfort and safety.

This research explores the acquisition and analysis of vibroacoustic signals generated during tissue-tool interactions, using a conventional aspiration needle enhanced with a proximally mounted MEMS audio sensor, to extract temperature information. Minimally invasive temperature monitoring is critical in thermotherapy applications, but current methods often rely on additional sensors or simulations of typical tissue behavior. In this study, a commercially available needle was inserted into water-saturated foams with temperatures ranging from 25°C to 55°C, varied in 5-degree increments. Given that temperature affects the speed of sound, water's heat capacity, and the mechanical properties of most tissues, it was hypothesized that the vibroacoustic signals recorded during needle insertion would carry temperature-dependent information. The acquired signals were segmented, processed, and analyzed using signal processing techniques and a deep learning algorithm. Results demonstrated that the audio signals contained distinct temperature-dependent features, enabling temperature prediction with a root mean squared error of approximately 3°C. We present these initial laboratory findings, highlighting significant potential for refinement. This novel approach could pave the way for a real-time, minimally invasive method for thermal monitoring in medical applications.

This paper delves into the inherent graphtopological structure utilized by algorithms such as LandMark in the context of RFID indoor positioning. We uncover the essence of these algorithms, which leverage the implicit topological relationships within signal features for message passing and positioning accuracy. Despite the theoretical advantages, practical applications of the LandMark algorithm have been hindered by issues related to signal propagation, the limitations of topological structures, neighbor tag selection, and simplistic weight distribution methods. To address these limitations, we propose a series of innovative improvements. Our approach includes data preprocessing techniques like B-spline interpolation and normalization to mitigate environmental noise and enhance signal integrity. We introduce the concept of spatiotemporal graphs that map signals into a high-dimensional space, allowing for the construction of dynamic graph structures that more accurately capture the temporal dynamics of signals. Furthermore, we employ the PNAConv algorithm, a graph neural network technique, to refine the message passing and feature aggregation process, optimizing the selection of neighboring tags. Our experiments, conducted across various datasets, demonstrate that our model maintains low error rates, showcasing its high precision and robustness in diverse environments. The results not only validate the effectiveness of our improved algorithm but also highlight the importance of understanding and exploiting the graph-topological structure inherent in signal-based positioning systems.

Given a picture classified as a Persian cat by AI, users may ask questions such as, "What are the contributions of the eyes and ears to the classification result?" or "Which features contribute the most?". Existing computational XAI methods have achieved success in explaining the AI output, but they cannot directly help users find the answers to such questions. In this paper, we propose a new approach that addresses the challenge by visually exploring XAI explanation through intuitive features. Our method computes the contributions of predefined semantic features (e.g., eyes, ears, body) to individual images and represents them with a quantitative representation. This approach provides an easily interpretable explanation of model classification and enables convenient visual exploration over multiple images. We also have developed a visualization prototype that allows users to efficiently explore, filter, and compare image groups based on the quantitative representation of the semantic features. We demonstrate the effectiveness of this approach in providing an intuitive and scalable way to interpret XAI results through a set of example scenarios and expert feedback.

Cloud data lakes are a modern way to manage vast amounts of data. They separate the compute and storage layers, which makes them highly scalable and cost-effective. However, query performance in cloud data lakes is relatively slow, and many attempts have been made to improve it in recent years. In this paper, we introduce our approach to this problem based on a novel caching technique. Our method relies on the observation that a significant bottleneck in query performance is caused by reading irrelevant files from remote storage. To address this, we suggest caching the relevant file names per query and reusing them in subsequent queries contained in some of the cached queries. Our solution uses ideas from both databases and computational geometry fields. Experiments based on the TPC-H benchmark demonstrate its feasibility and efficiency.