Mohit Tiwari

This reproducibility package supports an audited computational content analysis of cybersecurity-governance statements derived from seven official public sources. It contains 37 source-verified statements, precise locators, coding rules, an audit trail, executable Python and JavaScript workflows, generated tables and figures, and complete source–domain traceability. The audited analysis produced 26 exact agreements out of 37 statements (70.27%) and Cohen’s κ = 0.6452. The package supports a preliminary six-domain organizational maturity architecture and does not claim expert, construct, or institutional validation.

This reproducibility package supports an audited computational content analysis of cybersecurity-governance statements derived from seven official public sources. It contains 37 source-verified statements, precise locators, coding rules, an audit trail, executable Python and JavaScript workflows, generated tables and figures, and complete source–domain traceability. The audited analysis produced 26 exact agreements out of 37 statements (70.27%) and Cohen’s κ = 0.6452. The package supports a preliminary six-domain organizational maturity architecture and does not claim expert, construct, or institutional validation.

Frozen reproducibility evidence supporting the InCoWoCo 2026 conference manuscript ‘When Random Splits Mislead: Domain-Disjoint Evaluation of the PhiUSIIL Phishing URL Benchmark.’ The deposit is not a new release of the original PhiUSIIL dataset. It contains the frozen primary P1–P4 evaluation evidence plus separate authoritative Phase 7 feature-forensics and Phase 8 domain-disjoint feature-ablation archives. The study covers conservative canonical-URL deduplication; record-random and registered-domain-disjoint protocols; fixed-seed evaluation; prediction-derived metrics; bootstrap uncertainty and paired protocol comparisons; seen-versus-unseen-domain diagnostics; lexical sensitivity; feature forensics; and targeted ablation. The evidence demonstrates extreme internal PhiUSIIL benchmark separability under the evaluated controls. It does not establish target leakage, causality, memorization, or external, temporal, zero-day, campaign-level, or deployment generalization.

This paper introduces a novel framework named Reinforcement Learning-based Zero Trust with Trust Score (RLZT-TS) architecture for advanced dynamic trust management and security. This integrates an RL algorithm, a dynamic trust scoring (DTS) model, and a Zero Trust architecture (ZTA) in a hybrid cloud mesh infrastructure. This model can effectively detect cyber threats. An Advanced Multi-Factor Authentication (AMFA) protocol has been implemented in the security layer to secure the data during transmission over the cloud networks. The novel feature of RLZT-TS is a closed-loop adaptive trust model, where the trust score is continuously updated using RL algorithms, the access policy is dynamically adjusted based on the trust score, and the authentication requirements are dynamically changed based on the predicted risk level. The developed framework is analyzed through the MATLAB simulation, and outcomes illustrate the significant enhancement in trust score accuracy (98%), attack detection rate (99%), resource utilization (8.82%), policy adaptability (98%), and latency (8.2[Formula: see text]ms).

Effective leadership is a fundamental factor in the development of individuals, organizations, and societies. It involves the ability to establish a clear vision, inspire people, make informed decisions, manage resources, resolve conflicts, and achieve common objectives while maintaining ethical standards. Leadership is not confined to formal authority; it also involves influence, responsibility, empathy, and the ability to motivate others toward meaningful goals. Historical and contemporary leaders from India and other parts of the world demonstrate how leadership practices can influence social transformation, national development, organizational performance, and human welfare. Mahatma Gandhi, Dr. B. R. Ambedkar, Sardar Vallabhbhai Patel, A. P. J. Abdul Kalam, Nelson Mandela, Martin Luther King Jr., Abraham Lincoln, and Jacinda Ardern provide useful examples for examining different leadership qualities and approaches. This article discusses the meaning, characteristics, types, and significance of effective leadership, illustrating theoretical concepts through prominent leaders. It argues that effective leadership depends not on a single universal style but on the appropriate integration of vision, communication, adaptability, accountability, and ethical conduct.

Predictive artificial intelligence studies can report apparently strong performance while relying on predictors, transformations, or modelling procedures that would not be legitimate at the intended prediction point. Such problems include target-derived predictors, post-outcome information, temporally unavailable variables, inappropriate partitioning, duplicate or related observations across partitions, preprocessing before data separation, feature selection using evaluation data, and repeated use of a test set during model development. This protocol provides a domain-independent, reproducible procedure for auditing construct validity, prediction-time feature availability, feature admissibility, data leakage, partition integrity, preprocessing boundaries, model-selection practices, and robustness in supervised predictive-AI studies. It is applicable to classification and regression tasks in software engineering, healthcare, finance, cybersecurity, education, business analytics, and other data-driven disciplines. The protocol begins with a Prediction-Point Contract defining the target, unit of prediction, observation window, prediction point, outcome window, intended use, and legitimate information boundary. Every candidate predictor is then evaluated through a Feature Admissibility Record addressing construct meaning, provenance, temporal availability, target dependence, derivational dependence, and partition-specific learning. Pipeline-level risks are separately documented in a Leakage Register and Partition and Preprocessing Audit. Questionable predictors or modelling decisions are subjected to sensitivity and ablation analysis. The protocol concludes with a structured audit decision of PASS, CONDITIONAL, or FAIL. Required outputs are a Prediction-Point Contract, Feature Admissibility Matrix, Leakage Register, Partition and Preprocessing Audit, Sensitivity and Ablation Record, and Final Audit Checklist. The protocol is intended to make predictive-model validity claims more transparent, reproducible, auditable, and comparable across application domains. The protocol may be applied prospectively before model development, retrospectively to an existing study, or during peer review or replication.

Every day the need for Real-time Intelligence, Low-latency and Large-scale Data Processing capabilities for adaptive Public Safety Systems will become more important as the environment continues to change. In order to address this need, this research makes a proposal for an Edge-enhanced Smart Sensor System that combines IoT devices and Machine Learning Technologies to create intelligent safety systems. The overall architecture of the Edge-enhanced Smart Sensor System is a combination of IoT Sensors (Location-specific), Edge-based Data Preprocessing (Low-latency, Scalable, Efficient) and using a CNN-GRU Model for multimodal processing (text, audio and video). The coordination of Edge and Cloud systems will result in Low-latency, Scalable and Efficient Processing Solutions to provide an Adaptive Intelligent System supporting Public Safety. The system is evaluated both in terms of detection performance and latency (96.4% and 285 ms, respectively), which are better than cloud-based systems. It also has consistent performance with growing sensor loads and noise, showing its resilience.

Abstract Network intrusion detection systems (NIDS) are commonly evaluated on fixed benchmark datasets, but high performance on clean data does not necessarily indicate robustness to controlled changes in network-traffic features. This study presents a transformation-specific robustness evaluation of machine-learning-based NIDS using the UNSW-NB15 dataset. Multilayer perceptron (MLP) and Random Forest (RF) classifiers were evaluated under controlled perturbations of flow duration, timing/jitter, and time-to-live (TTL) features. The perturbations preserve the original class labels and modify selected feature values by predefined magnitudes, allowing the sensitivity of detection models to be examined independently of changes in the underlying labels. Flow-duration and timing/jitter perturbations produced negligible changes in detection performance over the evaluated ranges, whereas negative TTL perturbations produced substantially larger degradation in threshold-dependent TPR for the MLP, demonstrating transformation-dependent robustness. To further investigate TTL sensitivity, we evaluated a TTL-aware SCARF training configuration against an Original SCARF baseline using the same three random seeds (42, 123, and 2026) and the same dTTL test conditions. On clean test data, Original SCARF achieved a TPR of 0.7585 ± 0.0291, AUROC of 0.8927 ± 0.0107, and AUPRC of 0.9171 ± 0.0101, while TTL-aware SCARF achieved a comparable TPR of 0.7596 ± 0.0308, AUROC of 0.8910 ± 0.0102, and AUPRC of 0.9154 ± 0.0094. Under negative dTTL perturbations, TTL-aware SCARF consistently showed lower TPR degradation than Original SCARF. The reduction in TPR degradation ranged from 36.7% to 39.7% across the evaluated negative dTTL levels. AUROC and AUPRC remained comparatively stable despite the observed changes in TPR, highlighting the distinction between threshold-dependent detection performance and overall ranking quality. Overall, the results demonstrate that robustness in NIDS is transformation- and model-dependent. The findings also show that improved clean predictive performance and robustness to controlled feature perturbations should be evaluated as separate properties. The TTL-aware SCARF configuration provides a consistent improvement in robustness under the specific controlled dTTL perturbations evaluated, while the results should not be interpreted as evidence of robustness to general packet-level network change

Artificial intelligence (AI) and machine learning (ML) are increasingly transforming forensic toxicology by improving the detection, classification, and interpretation of illicit substances. Conventional analytical techniques such as gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS/MS) remain essential tools in toxicological investigations; however, the growing complexity of toxicological datasets and the rapid emergence of novel psychoactive substances (NPS) require more advanced computational approaches. Machine learning algorithms including support vector machines, random forests, artificial neural networks, and deep learning models have demonstrated strong potential for analyzing high-dimensional spectral data, identifying structural analogues, predicting metabolic pathways, and supporting toxicological interpretation. These approaches enable more efficient pattern recognition in mass spectrometry and metabolomics data, facilitating faster identification of emerging substances and improving analytical sensitivity. This review examines recent developments in the application of machine learning in forensic toxicology, focusing on its role in illicit substance detection, structural classification, and predictive toxicology. Key challenges, including data availability, model interpretability, reproducibility, and legal admissibility of AI-generated evidence, are critically discussed. Future perspectives such as explainable AI, federated learning, and integration with high-throughput analytical platforms are also highlighted. Overall, the integration of machine learning with forensic analytical workflows offers significant opportunities to enhance the accuracy, efficiency, and reliability of toxicological investigations.

Abstract Higher-education institutions (HEIs) must reconcile broad cybersecurity frameworks with sector-specific governance, open academic environments, privacy obligations, research partnerships, and emerging artificial-intelligence (AI) risks. This study analyzes a publicly released 50-record crosswalk that maps ten proposed HEI cybersecurity-maturity domains to high-level components of ten authoritative cybersecurity, governance, maturity, incident-response, and responsible-AI frameworks. A structured content-analysis design was used. Each record was examined by maturity domain, source framework, and mapping strength (Direct, Supporting, or Contextual). Reproducible frequency, incidence, concentration, and diversity measures were calculated from the deposited CSV file. Of the 50 mappings, 36 (72%) were Direct, nine (18%) Supporting, and five (10%) Contextual. NIST Cybersecurity Framework 2.0 covered all ten proposed domains, while C2M2 v2.1 covered nine. The three most represented sources contributed 52% of records, but normalized source entropy was 0.904, indicating that the crosswalk was not dominated by a single framework. Incident preparedness and response had five Direct mappings; the AI-readiness domain had the most mixed profile and uniquely incorporated NIST AI RMF 1.0 and ISO/IEC 42001:2023. The findings support a layered HEI design in which a general risk framework provides the backbone, capability and control sources add operational depth, sector guidance supplies context, and AI standards extend the model. The analysis is descriptive and does not validate the domains, assess any institution, or establish maturity scores. Independent expert coding and empirical testing are required.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract isolates a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. All artifacts are publicly released under open licenses.

Every day the need for Real-time Intelligence, Low-latency and Large-scale Data Processing capabilities for adaptive Public Safety Systems will become more important as the environment continues to change. In order to address this need, this research makes a proposal for an Edge-enhanced Smart Sensor System that combines IoT devices and Machine Learning Technologies to create intelligent safety systems. The overall architecture of the Edge-enhanced Smart Sensor System is a combination of IoT Sensors (Location-specific), Edge-based Data Preprocessing (Low-latency, Scalable, Efficient) and using a CNN-GRU Model for multimodal processing (text, audio and video). The coordination of Edge and Cloud systems will result in Low-latency, Scalable and Efficient Processing Solutions to provide an Adaptive Intelligent System supporting Public Safety. The system is evaluated both in terms of detection performance and latency (96.4% and 285 ms, respectively), which are better than cloud-based systems. It also has consistent performance with growing sensor loads and noise, showing its resilience.

To create the best method of extraction for a blockchain similar to Bitcoin, this research applies reinforcement learning, or RL, via AI machine learning techniques. Without being aware of the specifics of a blockchain system architecture, the RL learning structure makes it possible to choose the best course of action. Honest extraction was formerly thought to be the most effective mining technique; however, it has now been shown that there are greater benefits to departing from fair mining. You may define the mining issue as a process of Markov decision making (MDP) and solve it to find the best mining technique. The values of many variables, which are hard to find and subject to change in the future, are necessary to solve the MDP. In order to continuously develop a mining approach with efficiency that is close to the ideal mining method, the research makes use of reinforcement learning. The extraction MDP problem's non-linear function objective is resolved using a novel multifunctional RL method. The multifunctional RL mining method may reach optimum performance across time-varying blockchain platforms, according to research results.

This protocol describes a sequential, multi-phase study to develop and validate an AI-Assisted Cyber Maturity Framework for Higher Education Institutions (AICMF-HEI). Existing cybersecurity assessments often inadequately represent the distributed governance, mixed missions, transient populations, research environments, third-party dependencies, and resource diversity of higher education institutions. Phase 1 will refine the construct and candidate domains through evidence synthesis. Phase 2 will use semi-structured interviews with higher-education and cybersecurity stakeholders. Phase 3 will establish content validity and consensus through a modified Delphi study. Phase 4 will develop and cognitively test an assessment instrument. Phase 5 will examine reliability and exploratory factor structure. Phase 6 will conduct confirmatory validation using an independent multi-institutional sample and evaluate inter-rater agreement, convergent and discriminant validity, measurement invariance, usability, and decision usefulness. Phase 7 will evaluate a bounded AI-assistance prototype for evidence retrieval, mapping, contradiction detection, and rationale drafting under human oversight. The protocol prespecifies governance, ethics, data protection, stopping criteria, and transparent reporting. It does not report empirical results. The intended outcome is an evidence-centred and managerially useful framework rather than an autonomous certification system.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts (SC01–SC05) covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract is intentionally minimal to isolate a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment, preserving detector identifiers, impact levels, and confidence metadata. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. Baseline static analysis results are reported for each vulnerability class. All artifacts are publicly released under open licenses.

Abstract Network intrusion detection systems (NIDS) are commonly evaluated on fixed benchmark datasets, but high performance on clean data does not necessarily indicate robustness to controlled changes in network-traffic features. This study presents a transformation-specific robustness evaluation of machine-learning-based NIDS using the UNSW-NB15 dataset. Multilayer perceptron (MLP) and Random Forest (RF) classifiers were evaluated under controlled perturbations of flow duration, timing/jitter, and time-to-live (TTL) features. The perturbations preserve the original class labels and modify selected feature values by predefined magnitudes, allowing the sensitivity of detection models to be examined independently of changes in the underlying labels. Flow-duration and timing/jitter perturbations produced negligible changes in detection performance over the evaluated ranges, whereas negative TTL perturbations produced substantially larger degradation in threshold-dependent TPR for the MLP, demonstrating transformation-dependent robustness. To further investigate TTL sensitivity, we evaluated a TTL-aware SCARF training configuration against an Original SCARF baseline using the same three random seeds (42, 123, and 2026) and the same dTTL test conditions. On clean test data, Original SCARF achieved a TPR of 0.7585 ± 0.0291, AUROC of 0.8927 ± 0.0107, and AUPRC of 0.9171 ± 0.0101, while TTL-aware SCARF achieved a comparable TPR of 0.7596 ± 0.0308, AUROC of 0.8910 ± 0.0102, and AUPRC of 0.9154 ± 0.0094. Under negative dTTL perturbations, TTL-aware SCARF consistently showed lower TPR degradation than Original SCARF. The reduction in TPR degradation ranged from 36.7% to 39.7% across the evaluated negative dTTL levels. AUROC and AUPRC remained comparatively stable despite the observed changes in TPR, highlighting the distinction between threshold-dependent detection performance and overall ranking quality. Overall, the results demonstrate that robustness in NIDS is transformation- and model-dependent. The findings also show that improved clean predictive performance and robustness to controlled feature perturbations should be evaluated as separate properties. The TTL-aware SCARF configuration provides a consistent improvement in robustness under the specific controlled dTTL perturbations evaluated, while the results should not be interpreted as evidence of robustness to general packet-level network change

With the rapid expansion of consumer electronics and the Internet of Things (IoT) ecosystem, ensuring secure communication in resource-constrained edge devices has become a critical challenge, especially under the threat of quantum attacks. Although lattice-based cryptographic schemes provide post-quantum security, they often suffer from large key storage and high computational overhead, limiting their applicability in edge environments. To address these challenges, we propose an identity-based linkable ring signcryption scheme constructed on the post quantum Lattice cryptography, tailored for edge-enabled IoT systems. The scheme employs a trapdoor generation algorithm to establish the master keys, utilizes compact Gaussian sampling to derive private keys for ring members, and integrates rejection sampling to generate user signatures with strong anonymity. Furthermore, a key encapsulation mechanism ensures identity protection during encryption. Security analysis under the Random Oracle Model (ROM) demonstrates the scheme's resistance to forgery and its linkability property. Performance evaluation shows that the proposed approach achieves smaller public keys, reduced encryption/decryption costs, and enhanced efficiency, making it well-suited for lightweight IoT and consumer electronics applications in the post-quantum era.

This dataset contains version 1.0.0 of the AICMF-HEI Research Instruments and Validation Codebook, developed to support the proposed development and validation of an AI-Assisted Cyber Maturity Framework for Higher Education Institutions.The package contains 90 prospective research-instrument items comprising 20 semi-structured interview prompts, 30 modified-Delphi rating items, and 40 pilot-survey capability items. It also provides provisional decision rules, maturity anchors, variable definitions, response formats, analysis guidance, provenance documentation, and quality-control information.These instruments are intended for future expert review, cognitive testing, Delphi refinement, pilot administration, and psychometric validation. They are proposed and unvalidated and must not be represented as an empirically validated measurement instrument. The release contains no participant responses, institutional maturity scores, identifiable personal information, confidential cybersecurity evidence, or fabricated empirical results.Any future participant-based use requires appropriate institutional ethics approval, informed consent, secure data handling, anonymisation, and disclosure-risk assessment. AI assistance must remain bounded and human governed; responsibility for interpreting evidence and assigning maturity judgements remains with qualified human researchers and institutional reviewers.Generative AI assisted with language refinement, schema prototyping, and file preparation. The creator defined the research problem, reviewed the dataset structure and source metadata, and remains responsible for its accuracy, originality, and future validation.

Abstract Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts (SC01–SC05) covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract is intentionally minimal to isolate a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment, preserving detector identifiers, impact levels, and confidence metadata. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. Baseline static analysis results are reported for each vulnerability class. All artifacts are publicly released under open licenses.

Artificial intelligence (AI) and machine learning (ML) are increasingly transforming forensic toxicology by improving the detection, classification, and interpretation of illicit substances. Conventional analytical techniques such as gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS/MS) remain essential tools in toxicological investigations; however, the growing complexity of toxicological datasets and the rapid emergence of novel psychoactive substances (NPS) require more advanced computational approaches. Machine learning algorithms including support vector machines, random forests, artificial neural networks, and deep learning models have demonstrated strong potential for analyzing high-dimensional spectral data, identifying structural analogues, predicting metabolic pathways, and supporting toxicological interpretation. These approaches enable more efficient pattern recognition in mass spectrometry and metabolomics data, facilitating faster identification of emerging substances and improving analytical sensitivity. This review examines recent developments in the application of machine learning in forensic toxicology, focusing on its role in illicit substance detection, structural classification, and predictive toxicology. Key challenges, including data availability, model interpretability, reproducibility, and legal admissibility of AI-generated evidence, are critically discussed. Future perspectives such as explainable AI, federated learning, and integration with high-throughput analytical platforms are also highlighted. Overall, the integration of machine learning with forensic analytical workflows offers significant opportunities to enhance the accuracy, efficiency, and reliability of toxicological investigations.

Early detection of heart disease is essential for minimizing mortality, particularly in resource-constrained health-care environments. Recent models have shown accurate performance for medical diagnosis; however, they often struggles from limited interpretability and sensitivity to missing clinical data. To address these challenges, this paper proposes A Label-Enhanced Graph Neural Network Optimized by Elk Herd Optimization for Heart Disease Detection (HDD–LEGNN–EHOA). The input data is taken from Cleveland Heart disease data collection. Initially, data are pre-processed using Generalized Correntropy Sparse Gauss–Hermite Quadrature Filter (GCSGQF) to impute missing values. Then, the processed data is given to Secretary Bird Optimization Algorithm (SBOA) to select optimal characteristics. Then, the selected attributes fed to Label-Enhanced Graph Neural Network (LEGNN) to detect heart disease as heart disease present and heart disease absent. Elk Herd Optimization Algorithm (EHOA) to fine-tune the weight parameter of LEGNN. The newly suggested HDD–LEGNN–EHOA framework is assessed based on some metrics like accuracy, sensitivity, and computational time. Finally, the performance of HDD–LEGNN–EHOA method provides attains 25.68%, 22.54%, and 31.24% higher accuracy when compared to the existing models, respectively.

This dataset presents version 1.0.0 of the HEI Cybersecurity Governance and Policy Evidence Catalogue. It contains 30 curated records describing authoritative and sector-relevant public resources for designing, evaluating, and auditing cybersecurity governance in higher education institutions. The catalogue covers governance, policy, risk management, privacy, institutional resilience, workforce capability, supplier governance, incident preparedness, and responsible artificial intelligence. Each record provides structured source metadata and an original conceptual mapping to higher-education cybersecurity governance and evidence requirements. The dataset contains no participant responses, confidential institutional records, cybersecurity findings, maturity scores, or fabricated empirical results. The mappings represent the author’s conceptual synthesis and require expert review and future empirical validation. No proprietary standards text is reproduced.

The combination of blockchain with the application of AI results from the growing need for data safety and confidentiality in the healthcare industry. The article is concerned with the integration of these systems in the clinical diagnostics, health records, and pharmacological supply chain leadership. To investigate how blockchain guarantees data durability via transferred storage methods and how AI uses this data for making decision systems by means of deep learning and immediate analysis, it reviews the literature and analyses case studies. The findings demonstrate that although AI improves diagnostic accuracy and care decision-making, blockchain distributes data for patient confidentiality and integrity. There is a lot of promise for intelligent and individualised growth in the healthcare sector with the combination of blockchain and AI technologies.

Frozen reproducibility evidence supporting the InCoWoCo 2026 conference manuscript ‘When Random Splits Mislead: Domain-Disjoint Evaluation of the PhiUSIIL Phishing URL Benchmark.’ The deposit is not a new release of the original PhiUSIIL dataset. It contains the frozen primary P1–P4 evaluation evidence plus separate authoritative Phase 7 feature-forensics and Phase 8 domain-disjoint feature-ablation archives. The study covers conservative canonical-URL deduplication; record-random and registered-domain-disjoint protocols; fixed-seed evaluation; prediction-derived metrics; bootstrap uncertainty and paired protocol comparisons; seen-versus-unseen-domain diagnostics; lexical sensitivity; feature forensics; and targeted ablation. The evidence demonstrates extreme internal PhiUSIIL benchmark separability under the evaluated controls. It does not establish target leakage, causality, memorization, or external, temporal, zero-day, campaign-level, or deployment generalization.

Early detection of heart disease is essential for minimizing mortality, particularly in resource-constrained health-care environments. Recent models have shown accurate performance for medical diagnosis; however, they often struggles from limited interpretability and sensitivity to missing clinical data. To address these challenges, this paper proposes A Label-Enhanced Graph Neural Network Optimized by Elk Herd Optimization for Heart Disease Detection (HDD–LEGNN–EHOA). The input data is taken from Cleveland Heart disease data collection. Initially, data are pre-processed using Generalized Correntropy Sparse Gauss–Hermite Quadrature Filter (GCSGQF) to impute missing values. Then, the processed data is given to Secretary Bird Optimization Algorithm (SBOA) to select optimal characteristics. Then, the selected attributes fed to Label-Enhanced Graph Neural Network (LEGNN) to detect heart disease as heart disease present and heart disease absent. Elk Herd Optimization Algorithm (EHOA) to fine-tune the weight parameter of LEGNN. The newly suggested HDD–LEGNN–EHOA framework is assessed based on some metrics like accuracy, sensitivity, and computational time. Finally, the performance of HDD–LEGNN–EHOA method provides attains 25.68%, 22.54%, and 31.24% higher accuracy when compared to the existing models, respectively.

Abstract Higher-education institutions (HEIs) must reconcile broad cybersecurity frameworks with sector-specific governance, open academic environments, privacy obligations, research partnerships, and emerging artificial-intelligence (AI) risks. This study analyzes a publicly released 50-record crosswalk that maps ten proposed HEI cybersecurity-maturity domains to high-level components of ten authoritative cybersecurity, governance, maturity, incident-response, and responsible-AI frameworks. A structured content-analysis design was used. Each record was examined by maturity domain, source framework, and mapping strength (Direct, Supporting, or Contextual). Reproducible frequency, incidence, concentration, and diversity measures were calculated from the deposited CSV file. Of the 50 mappings, 36 (72%) were Direct, nine (18%) Supporting, and five (10%) Contextual. NIST Cybersecurity Framework 2.0 covered all ten proposed domains, while C2M2 v2.1 covered nine. The three most represented sources contributed 52% of records, but normalized source entropy was 0.904, indicating that the crosswalk was not dominated by a single framework. Incident preparedness and response had five Direct mappings; the AI-readiness domain had the most mixed profile and uniquely incorporated NIST AI RMF 1.0 and ISO/IEC 42001:2023. The findings support a layered HEI design in which a general risk framework provides the backbone, capability and control sources add operational depth, sector guidance supplies context, and AI standards extend the model. The analysis is descriptive and does not validate the domains, assess any institution, or establish maturity scores. Independent expert coding and empirical testing are required.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract isolates a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. All artifacts are publicly released under open licenses.

The smart-city systems of today require computing systems that can manage large volumes of latency-sensitive IoT data with a very small power cost. Nevertheless, the dynamics and distribution of such conditions do not go well with traditional cloud-centric infrastructures and, as a result, can cause network congestion, slow analytics, The research suggests a Cloud-Edge-Fog Hybrid (CEFH) system to combine real-time reinforcement learning to coordinate tasks among levels on intelligence. The system is constantly monitoring resource availability, network conditions and energy to assign the computational tasks dynamically. The use of experimental results on a heterogeneous IoT testbed has shown a reduction of latency by 65 percent, an energy saving of 41 percent and an increase of throughput by 23 percent compared to recent hybrid models. The system was stable and stood the test of high workload conditions, and a high rate of task success of 96.7, which is indicative of scalability and stability.

Predictive artificial intelligence studies can report apparently strong performance while relying on predictors, transformations, or modelling procedures that would not be legitimate at the intended prediction point. Such problems include target-derived predictors, post-outcome information, temporally unavailable variables, inappropriate partitioning, duplicate or related observations across partitions, preprocessing before data separation, feature selection using evaluation data, and repeated use of a test set during model development. This protocol provides a domain-independent, reproducible procedure for auditing construct validity, prediction-time feature availability, feature admissibility, data leakage, partition integrity, preprocessing boundaries, model-selection practices, and robustness in supervised predictive-AI studies. It is applicable to classification and regression tasks in software engineering, healthcare, finance, cybersecurity, education, business analytics, and other data-driven disciplines. The protocol begins with a Prediction-Point Contract defining the target, unit of prediction, observation window, prediction point, outcome window, intended use, and legitimate information boundary. Every candidate predictor is then evaluated through a Feature Admissibility Record addressing construct meaning, provenance, temporal availability, target dependence, derivational dependence, and partition-specific learning. Pipeline-level risks are separately documented in a Leakage Register and Partition and Preprocessing Audit. Questionable predictors or modelling decisions are subjected to sensitivity and ablation analysis. The protocol concludes with a structured audit decision of PASS, CONDITIONAL, or FAIL. Required outputs are a Prediction-Point Contract, Feature Admissibility Matrix, Leakage Register, Partition and Preprocessing Audit, Sensitivity and Ablation Record, and Final Audit Checklist. The protocol is intended to make predictive-model validity claims more transparent, reproducible, auditable, and comparable across application domains. The protocol may be applied prospectively before model development, retrospectively to an existing study, or during peer review or replication.

This dataset contains version 1.0.0 of the AICMF-HEI Research Instruments and Validation Codebook, developed to support the proposed development and validation of an AI-Assisted Cyber Maturity Framework for Higher Education Institutions.The package contains 90 prospective research-instrument items comprising 20 semi-structured interview prompts, 30 modified-Delphi rating items, and 40 pilot-survey capability items. It also provides provisional decision rules, maturity anchors, variable definitions, response formats, analysis guidance, provenance documentation, and quality-control information.These instruments are intended for future expert review, cognitive testing, Delphi refinement, pilot administration, and psychometric validation. They are proposed and unvalidated and must not be represented as an empirically validated measurement instrument. The release contains no participant responses, institutional maturity scores, identifiable personal information, confidential cybersecurity evidence, or fabricated empirical results.Any future participant-based use requires appropriate institutional ethics approval, informed consent, secure data handling, anonymisation, and disclosure-risk assessment. AI assistance must remain bounded and human governed; responsibility for interpreting evidence and assigning maturity judgements remains with qualified human researchers and institutional reviewers.Generative AI assisted with language refinement, schema prototyping, and file preparation. The creator defined the research problem, reviewed the dataset structure and source metadata, and remains responsible for its accuracy, originality, and future validation.

Abstract Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts (SC01–SC05) covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract is intentionally minimal to isolate a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment, preserving detector identifiers, impact levels, and confidence metadata. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. Baseline static analysis results are reported for each vulnerability class. All artifacts are publicly released under open licenses.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets for fair evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts covering reentrancy (SWC-107), integer overflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract isolates a single vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. All artifacts are publicly released under open licenses.

This dataset presents version 1.0.0 of the HEI Cybersecurity Governance and Policy Evidence Catalogue. It contains 30 curated records describing authoritative and sector-relevant public resources for designing, evaluating, and auditing cybersecurity governance in higher education institutions. The catalogue covers governance, policy, risk management, privacy, institutional resilience, workforce capability, supplier governance, incident preparedness, and responsible artificial intelligence. Each record provides structured source metadata and an original conceptual mapping to higher-education cybersecurity governance and evidence requirements. The dataset contains no participant responses, confidential institutional records, cybersecurity findings, maturity scores, or fabricated empirical results. The mappings represent the author’s conceptual synthesis and require expert review and future empirical validation. No proprietary standards text is reproduced.

The combination of blockchain with the application of AI results from the growing need for data safety and confidentiality in the healthcare industry. The article is concerned with the integration of these systems in the clinical diagnostics, health records, and pharmacological supply chain leadership. To investigate how blockchain guarantees data durability via transferred storage methods and how AI uses this data for making decision systems by means of deep learning and immediate analysis, it reviews the literature and analyses case studies. The findings demonstrate that although AI improves diagnostic accuracy and care decision-making, blockchain distributes data for patient confidentiality and integrity. There is a lot of promise for intelligent and individualised growth in the healthcare sector with the combination of blockchain and AI technologies.

To create the best method of extraction for a blockchain similar to Bitcoin, this research applies reinforcement learning, or RL, via AI machine learning techniques. Without being aware of the specifics of a blockchain system architecture, the RL learning structure makes it possible to choose the best course of action. Honest extraction was formerly thought to be the most effective mining technique; however, it has now been shown that there are greater benefits to departing from fair mining. You may define the mining issue as a process of Markov decision making (MDP) and solve it to find the best mining technique. The values of many variables, which are hard to find and subject to change in the future, are necessary to solve the MDP. In order to continuously develop a mining approach with efficiency that is close to the ideal mining method, the research makes use of reinforcement learning. The extraction MDP problem's non-linear function objective is resolved using a novel multifunctional RL method. The multifunctional RL mining method may reach optimum performance across time-varying blockchain platforms, according to research results.

This paper introduces a novel framework named Reinforcement Learning-based Zero Trust with Trust Score (RLZT-TS) architecture for advanced dynamic trust management and security. This integrates an RL algorithm, a dynamic trust scoring (DTS) model, and a Zero Trust architecture (ZTA) in a hybrid cloud mesh infrastructure. This model can effectively detect cyber threats. An Advanced Multi-Factor Authentication (AMFA) protocol has been implemented in the security layer to secure the data during transmission over the cloud networks. The novel feature of RLZT-TS is a closed-loop adaptive trust model, where the trust score is continuously updated using RL algorithms, the access policy is dynamically adjusted based on the trust score, and the authentication requirements are dynamically changed based on the predicted risk level. The developed framework is analyzed through the MATLAB simulation, and outcomes illustrate the significant enhancement in trust score accuracy (98%), attack detection rate (99%), resource utilization (8.82%), policy adaptability (98%), and latency (8.2[Formula: see text]ms).

The smart-city systems of today require computing systems that can manage large volumes of latency-sensitive IoT data with a very small power cost. Nevertheless, the dynamics and distribution of such conditions do not go well with traditional cloud-centric infrastructures and, as a result, can cause network congestion, slow analytics, The research suggests a Cloud-Edge-Fog Hybrid (CEFH) system to combine real-time reinforcement learning to coordinate tasks among levels on intelligence. The system is constantly monitoring resource availability, network conditions and energy to assign the computational tasks dynamically. The use of experimental results on a heterogeneous IoT testbed has shown a reduction of latency by 65 percent, an energy saving of 41 percent and an increase of throughput by 23 percent compared to recent hybrid models. The system was stable and stood the test of high workload conditions, and a high rate of task success of 96.7, which is indicative of scalability and stability.

With the rapid expansion of consumer electronics and the Internet of Things (IoT) ecosystem, ensuring secure communication in resource-constrained edge devices has become a critical challenge, especially under the threat of quantum attacks. Although lattice-based cryptographic schemes provide post-quantum security, they often suffer from large key storage and high computational overhead, limiting their applicability in edge environments. To address these challenges, we propose an identity-based linkable ring signcryption scheme constructed on the post quantum Lattice cryptography, tailored for edge-enabled IoT systems. The scheme employs a trapdoor generation algorithm to establish the master keys, utilizes compact Gaussian sampling to derive private keys for ring members, and integrates rejection sampling to generate user signatures with strong anonymity. Furthermore, a key encapsulation mechanism ensures identity protection during encryption. Security analysis under the Random Oracle Model (ROM) demonstrates the scheme's resistance to forgery and its linkability property. Performance evaluation shows that the proposed approach achieves smaller public keys, reduced encryption/decryption costs, and enhanced efficiency, making it well-suited for lightweight IoT and consumer electronics applications in the post-quantum era.

This reproducibility package supports an audited computational content analysis of cybersecurity-governance statements derived from seven official public sources. It contains 37 source-verified statements, precise locators, coding rules, an audit trail, executable Python and JavaScript workflows, generated tables and figures, and complete source–domain traceability. The audited analysis produced 26 exact agreements out of 37 statements (70.27%) and Cohen’s κ = 0.6452. The package supports a preliminary six-domain organizational maturity architecture and does not claim expert, construct, or institutional validation.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets for fair evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts covering reentrancy (SWC-107), integer overflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract isolates a single vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. All artifacts are publicly released under open licenses.

Effective leadership is a fundamental factor in the development of individuals, organizations, and societies. It involves the ability to establish a clear vision, inspire people, make informed decisions, manage resources, resolve conflicts, and achieve common objectives while maintaining ethical standards. Leadership is not confined to formal authority; it also involves influence, responsibility, empathy, and the ability to motivate others toward meaningful goals. Historical and contemporary leaders from India and other parts of the world demonstrate how leadership practices can influence social transformation, national development, organizational performance, and human welfare. Mahatma Gandhi, Dr. B. R. Ambedkar, Sardar Vallabhbhai Patel, A. P. J. Abdul Kalam, Nelson Mandela, Martin Luther King Jr., Abraham Lincoln, and Jacinda Ardern provide useful examples for examining different leadership qualities and approaches. This article discusses the meaning, characteristics, types, and significance of effective leadership, illustrating theoretical concepts through prominent leaders. It argues that effective leadership depends not on a single universal style but on the appropriate integration of vision, communication, adaptability, accountability, and ethical conduct.

This protocol describes a sequential, multi-phase study to develop and validate an AI-Assisted Cyber Maturity Framework for Higher Education Institutions (AICMF-HEI). Existing cybersecurity assessments often inadequately represent the distributed governance, mixed missions, transient populations, research environments, third-party dependencies, and resource diversity of higher education institutions. Phase 1 will refine the construct and candidate domains through evidence synthesis. Phase 2 will use semi-structured interviews with higher-education and cybersecurity stakeholders. Phase 3 will establish content validity and consensus through a modified Delphi study. Phase 4 will develop and cognitively test an assessment instrument. Phase 5 will examine reliability and exploratory factor structure. Phase 6 will conduct confirmatory validation using an independent multi-institutional sample and evaluate inter-rater agreement, convergent and discriminant validity, measurement invariance, usability, and decision usefulness. Phase 7 will evaluate a bounded AI-assistance prototype for evidence retrieval, mapping, contradiction detection, and rationale drafting under human oversight. The protocol prespecifies governance, ethics, data protection, stopping criteria, and transparent reporting. It does not report empirical results. The intended outcome is an evidence-centred and managerially useful framework rather than an autonomous certification system.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts (SC01–SC05) covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract is intentionally minimal to isolate a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment, preserving detector identifiers, impact levels, and confidence metadata. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. Baseline static analysis results are reported for each vulnerability class. All artifacts are publicly released under open licenses.

Effective leadership is a fundamental factor in the development of individuals, organizations, and societies. It involves the ability to establish a clear vision, inspire people, make informed decisions, manage resources, resolve conflicts, and achieve common objectives while maintaining ethical standards. Leadership is not confined to formal authority; it also involves influence, responsibility, empathy, and the ability to motivate others toward meaningful goals. Historical and contemporary leaders from India and other parts of the world demonstrate how leadership practices can influence social transformation, national development, organizational performance, and human welfare. Mahatma Gandhi, Dr. B. R. Ambedkar, Sardar Vallabhbhai Patel, A. P. J. Abdul Kalam, Nelson Mandela, Martin Luther King Jr., Abraham Lincoln, and Jacinda Ardern provide useful examples for examining different leadership qualities and approaches. This article discusses the meaning, characteristics, types, and significance of effective leadership, illustrating theoretical concepts through prominent leaders. It argues that effective leadership depends not on a single universal style but on the appropriate integration of vision, communication, adaptability, accountability, and ethical conduct.

Predictive artificial intelligence studies can report apparently strong performance while relying on predictors, transformations, or modelling procedures that would not be legitimate at the intended prediction point. Such problems include target-derived predictors, post-outcome information, temporally unavailable variables, inappropriate partitioning, duplicate or related observations across partitions, preprocessing before data separation, feature selection using evaluation data, and repeated use of a test set during model development. This protocol provides a domain-independent, reproducible procedure for auditing construct validity, prediction-time feature availability, feature admissibility, data leakage, partition integrity, preprocessing boundaries, model-selection practices, and robustness in supervised predictive-AI studies. It is applicable to classification and regression tasks in software engineering, healthcare, finance, cybersecurity, education, business analytics, and other data-driven disciplines. The protocol begins with a Prediction-Point Contract defining the target, unit of prediction, observation window, prediction point, outcome window, intended use, and legitimate information boundary. Every candidate predictor is then evaluated through a Feature Admissibility Record addressing construct meaning, provenance, temporal availability, target dependence, derivational dependence, and partition-specific learning. Pipeline-level risks are separately documented in a Leakage Register and Partition and Preprocessing Audit. Questionable predictors or modelling decisions are subjected to sensitivity and ablation analysis. The protocol concludes with a structured audit decision of PASS, CONDITIONAL, or FAIL. Required outputs are a Prediction-Point Contract, Feature Admissibility Matrix, Leakage Register, Partition and Preprocessing Audit, Sensitivity and Ablation Record, and Final Audit Checklist. The protocol is intended to make predictive-model validity claims more transparent, reproducible, auditable, and comparable across application domains. The protocol may be applied prospectively before model development, retrospectively to an existing study, or during peer review or replication.

Every day the need for Real-time Intelligence, Low-latency and Large-scale Data Processing capabilities for adaptive Public Safety Systems will become more important as the environment continues to change. In order to address this need, this research makes a proposal for an Edge-enhanced Smart Sensor System that combines IoT devices and Machine Learning Technologies to create intelligent safety systems. The overall architecture of the Edge-enhanced Smart Sensor System is a combination of IoT Sensors (Location-specific), Edge-based Data Preprocessing (Low-latency, Scalable, Efficient) and using a CNN-GRU Model for multimodal processing (text, audio and video). The coordination of Edge and Cloud systems will result in Low-latency, Scalable and Efficient Processing Solutions to provide an Adaptive Intelligent System supporting Public Safety. The system is evaluated both in terms of detection performance and latency (96.4% and 285 ms, respectively), which are better than cloud-based systems. It also has consistent performance with growing sensor loads and noise, showing its resilience.

Abstract Network intrusion detection systems (NIDS) are commonly evaluated on fixed benchmark datasets, but high performance on clean data does not necessarily indicate robustness to controlled changes in network-traffic features. This study presents a transformation-specific robustness evaluation of machine-learning-based NIDS using the UNSW-NB15 dataset. Multilayer perceptron (MLP) and Random Forest (RF) classifiers were evaluated under controlled perturbations of flow duration, timing/jitter, and time-to-live (TTL) features. The perturbations preserve the original class labels and modify selected feature values by predefined magnitudes, allowing the sensitivity of detection models to be examined independently of changes in the underlying labels. Flow-duration and timing/jitter perturbations produced negligible changes in detection performance over the evaluated ranges, whereas negative TTL perturbations produced substantially larger degradation in threshold-dependent TPR for the MLP, demonstrating transformation-dependent robustness. To further investigate TTL sensitivity, we evaluated a TTL-aware SCARF training configuration against an Original SCARF baseline using the same three random seeds (42, 123, and 2026) and the same dTTL test conditions. On clean test data, Original SCARF achieved a TPR of 0.7585 ± 0.0291, AUROC of 0.8927 ± 0.0107, and AUPRC of 0.9171 ± 0.0101, while TTL-aware SCARF achieved a comparable TPR of 0.7596 ± 0.0308, AUROC of 0.8910 ± 0.0102, and AUPRC of 0.9154 ± 0.0094. Under negative dTTL perturbations, TTL-aware SCARF consistently showed lower TPR degradation than Original SCARF. The reduction in TPR degradation ranged from 36.7% to 39.7% across the evaluated negative dTTL levels. AUROC and AUPRC remained comparatively stable despite the observed changes in TPR, highlighting the distinction between threshold-dependent detection performance and overall ranking quality. Overall, the results demonstrate that robustness in NIDS is transformation- and model-dependent. The findings also show that improved clean predictive performance and robustness to controlled feature perturbations should be evaluated as separate properties. The TTL-aware SCARF configuration provides a consistent improvement in robustness under the specific controlled dTTL perturbations evaluated, while the results should not be interpreted as evidence of robustness to general packet-level network change

Artificial intelligence (AI) and machine learning (ML) are increasingly transforming forensic toxicology by improving the detection, classification, and interpretation of illicit substances. Conventional analytical techniques such as gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS/MS) remain essential tools in toxicological investigations; however, the growing complexity of toxicological datasets and the rapid emergence of novel psychoactive substances (NPS) require more advanced computational approaches. Machine learning algorithms including support vector machines, random forests, artificial neural networks, and deep learning models have demonstrated strong potential for analyzing high-dimensional spectral data, identifying structural analogues, predicting metabolic pathways, and supporting toxicological interpretation. These approaches enable more efficient pattern recognition in mass spectrometry and metabolomics data, facilitating faster identification of emerging substances and improving analytical sensitivity. This review examines recent developments in the application of machine learning in forensic toxicology, focusing on its role in illicit substance detection, structural classification, and predictive toxicology. Key challenges, including data availability, model interpretability, reproducibility, and legal admissibility of AI-generated evidence, are critically discussed. Future perspectives such as explainable AI, federated learning, and integration with high-throughput analytical platforms are also highlighted. Overall, the integration of machine learning with forensic analytical workflows offers significant opportunities to enhance the accuracy, efficiency, and reliability of toxicological investigations.

Frozen reproducibility evidence supporting the InCoWoCo 2026 conference manuscript ‘When Random Splits Mislead: Domain-Disjoint Evaluation of the PhiUSIIL Phishing URL Benchmark.’ The deposit is not a new release of the original PhiUSIIL dataset. It contains the frozen primary P1–P4 evaluation evidence plus separate authoritative Phase 7 feature-forensics and Phase 8 domain-disjoint feature-ablation archives. The study covers conservative canonical-URL deduplication; record-random and registered-domain-disjoint protocols; fixed-seed evaluation; prediction-derived metrics; bootstrap uncertainty and paired protocol comparisons; seen-versus-unseen-domain diagnostics; lexical sensitivity; feature forensics; and targeted ablation. The evidence demonstrates extreme internal PhiUSIIL benchmark separability under the evaluated controls. It does not establish target leakage, causality, memorization, or external, temporal, zero-day, campaign-level, or deployment generalization.

This paper introduces a novel framework named Reinforcement Learning-based Zero Trust with Trust Score (RLZT-TS) architecture for advanced dynamic trust management and security. This integrates an RL algorithm, a dynamic trust scoring (DTS) model, and a Zero Trust architecture (ZTA) in a hybrid cloud mesh infrastructure. This model can effectively detect cyber threats. An Advanced Multi-Factor Authentication (AMFA) protocol has been implemented in the security layer to secure the data during transmission over the cloud networks. The novel feature of RLZT-TS is a closed-loop adaptive trust model, where the trust score is continuously updated using RL algorithms, the access policy is dynamically adjusted based on the trust score, and the authentication requirements are dynamically changed based on the predicted risk level. The developed framework is analyzed through the MATLAB simulation, and outcomes illustrate the significant enhancement in trust score accuracy (98%), attack detection rate (99%), resource utilization (8.82%), policy adaptability (98%), and latency (8.2[Formula: see text]ms).

Abstract Higher-education institutions (HEIs) must reconcile broad cybersecurity frameworks with sector-specific governance, open academic environments, privacy obligations, research partnerships, and emerging artificial-intelligence (AI) risks. This study analyzes a publicly released 50-record crosswalk that maps ten proposed HEI cybersecurity-maturity domains to high-level components of ten authoritative cybersecurity, governance, maturity, incident-response, and responsible-AI frameworks. A structured content-analysis design was used. Each record was examined by maturity domain, source framework, and mapping strength (Direct, Supporting, or Contextual). Reproducible frequency, incidence, concentration, and diversity measures were calculated from the deposited CSV file. Of the 50 mappings, 36 (72%) were Direct, nine (18%) Supporting, and five (10%) Contextual. NIST Cybersecurity Framework 2.0 covered all ten proposed domains, while C2M2 v2.1 covered nine. The three most represented sources contributed 52% of records, but normalized source entropy was 0.904, indicating that the crosswalk was not dominated by a single framework. Incident preparedness and response had five Direct mappings; the AI-readiness domain had the most mixed profile and uniquely incorporated NIST AI RMF 1.0 and ISO/IEC 42001:2023. The findings support a layered HEI design in which a general risk framework provides the backbone, capability and control sources add operational depth, sector guidance supplies context, and AI standards extend the model. The analysis is descriptive and does not validate the domains, assess any institution, or establish maturity scores. Independent expert coding and empirical testing are required.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract isolates a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. All artifacts are publicly released under open licenses.

This protocol describes a sequential, multi-phase study to develop and validate an AI-Assisted Cyber Maturity Framework for Higher Education Institutions (AICMF-HEI). Existing cybersecurity assessments often inadequately represent the distributed governance, mixed missions, transient populations, research environments, third-party dependencies, and resource diversity of higher education institutions. Phase 1 will refine the construct and candidate domains through evidence synthesis. Phase 2 will use semi-structured interviews with higher-education and cybersecurity stakeholders. Phase 3 will establish content validity and consensus through a modified Delphi study. Phase 4 will develop and cognitively test an assessment instrument. Phase 5 will examine reliability and exploratory factor structure. Phase 6 will conduct confirmatory validation using an independent multi-institutional sample and evaluate inter-rater agreement, convergent and discriminant validity, measurement invariance, usability, and decision usefulness. Phase 7 will evaluate a bounded AI-assistance prototype for evidence retrieval, mapping, contradiction detection, and rationale drafting under human oversight. The protocol prespecifies governance, ethics, data protection, stopping criteria, and transparent reporting. It does not report empirical results. The intended outcome is an evidence-centred and managerially useful framework rather than an autonomous certification system.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts (SC01–SC05) covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract is intentionally minimal to isolate a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment, preserving detector identifiers, impact levels, and confidence metadata. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. Baseline static analysis results are reported for each vulnerability class. All artifacts are publicly released under open licenses.

This dataset contains version 1.0.0 of the AICMF-HEI Research Instruments and Validation Codebook, developed to support the proposed development and validation of an AI-Assisted Cyber Maturity Framework for Higher Education Institutions.The package contains 90 prospective research-instrument items comprising 20 semi-structured interview prompts, 30 modified-Delphi rating items, and 40 pilot-survey capability items. It also provides provisional decision rules, maturity anchors, variable definitions, response formats, analysis guidance, provenance documentation, and quality-control information.These instruments are intended for future expert review, cognitive testing, Delphi refinement, pilot administration, and psychometric validation. They are proposed and unvalidated and must not be represented as an empirically validated measurement instrument. The release contains no participant responses, institutional maturity scores, identifiable personal information, confidential cybersecurity evidence, or fabricated empirical results.Any future participant-based use requires appropriate institutional ethics approval, informed consent, secure data handling, anonymisation, and disclosure-risk assessment. AI assistance must remain bounded and human governed; responsibility for interpreting evidence and assigning maturity judgements remains with qualified human researchers and institutional reviewers.Generative AI assisted with language refinement, schema prototyping, and file preparation. The creator defined the research problem, reviewed the dataset structure and source metadata, and remains responsible for its accuracy, originality, and future validation.

Abstract Smart contract vulnerability benchmarking lacks standardized, reproducible datasets that enable fair and consistent evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts (SC01–SC05) covering reentrancy (SWC-107), integer overflow/underflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract is intentionally minimal to isolate a single structural vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment, preserving detector identifiers, impact levels, and confidence metadata. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. Baseline static analysis results are reported for each vulnerability class. All artifacts are publicly released under open licenses.

This dataset presents version 1.0.0 of the HEI Cybersecurity Governance and Policy Evidence Catalogue. It contains 30 curated records describing authoritative and sector-relevant public resources for designing, evaluating, and auditing cybersecurity governance in higher education institutions. The catalogue covers governance, policy, risk management, privacy, institutional resilience, workforce capability, supplier governance, incident preparedness, and responsible artificial intelligence. Each record provides structured source metadata and an original conceptual mapping to higher-education cybersecurity governance and evidence requirements. The dataset contains no participant responses, confidential institutional records, cybersecurity findings, maturity scores, or fabricated empirical results. The mappings represent the author’s conceptual synthesis and require expert review and future empirical validation. No proprietary standards text is reproduced.

The combination of blockchain with the application of AI results from the growing need for data safety and confidentiality in the healthcare industry. The article is concerned with the integration of these systems in the clinical diagnostics, health records, and pharmacological supply chain leadership. To investigate how blockchain guarantees data durability via transferred storage methods and how AI uses this data for making decision systems by means of deep learning and immediate analysis, it reviews the literature and analyses case studies. The findings demonstrate that although AI improves diagnostic accuracy and care decision-making, blockchain distributes data for patient confidentiality and integrity. There is a lot of promise for intelligent and individualised growth in the healthcare sector with the combination of blockchain and AI technologies.

Early detection of heart disease is essential for minimizing mortality, particularly in resource-constrained health-care environments. Recent models have shown accurate performance for medical diagnosis; however, they often struggles from limited interpretability and sensitivity to missing clinical data. To address these challenges, this paper proposes A Label-Enhanced Graph Neural Network Optimized by Elk Herd Optimization for Heart Disease Detection (HDD–LEGNN–EHOA). The input data is taken from Cleveland Heart disease data collection. Initially, data are pre-processed using Generalized Correntropy Sparse Gauss–Hermite Quadrature Filter (GCSGQF) to impute missing values. Then, the processed data is given to Secretary Bird Optimization Algorithm (SBOA) to select optimal characteristics. Then, the selected attributes fed to Label-Enhanced Graph Neural Network (LEGNN) to detect heart disease as heart disease present and heart disease absent. Elk Herd Optimization Algorithm (EHOA) to fine-tune the weight parameter of LEGNN. The newly suggested HDD–LEGNN–EHOA framework is assessed based on some metrics like accuracy, sensitivity, and computational time. Finally, the performance of HDD–LEGNN–EHOA method provides attains 25.68%, 22.54%, and 31.24% higher accuracy when compared to the existing models, respectively.

The smart-city systems of today require computing systems that can manage large volumes of latency-sensitive IoT data with a very small power cost. Nevertheless, the dynamics and distribution of such conditions do not go well with traditional cloud-centric infrastructures and, as a result, can cause network congestion, slow analytics, The research suggests a Cloud-Edge-Fog Hybrid (CEFH) system to combine real-time reinforcement learning to coordinate tasks among levels on intelligence. The system is constantly monitoring resource availability, network conditions and energy to assign the computational tasks dynamically. The use of experimental results on a heterogeneous IoT testbed has shown a reduction of latency by 65 percent, an energy saving of 41 percent and an increase of throughput by 23 percent compared to recent hybrid models. The system was stable and stood the test of high workload conditions, and a high rate of task success of 96.7, which is indicative of scalability and stability.

Smart contract vulnerability benchmarking lacks standardized, reproducible datasets for fair evaluation of static analysis tools. This paper presents SARMF-Bench, a compact, deterministic, and fully reproducible benchmark dataset comprising five SWC-aligned Solidity smart contracts covering reentrancy (SWC-107), integer overflow (SWC-101), access-control weakness (SWC-105), unchecked external calls (SWC-104), and denial-of-service via unbounded loops (SWC-113). Each contract isolates a single vulnerability pattern and is paired with machine-readable JSON outputs generated using Slither v0.11.5 in a version-locked environment. SARMF-Bench is archived across multiple open repositories with permanent DOIs to enable fully reproducible smart contract security tool evaluation experiments. All artifacts are publicly released under open licenses.

To create the best method of extraction for a blockchain similar to Bitcoin, this research applies reinforcement learning, or RL, via AI machine learning techniques. Without being aware of the specifics of a blockchain system architecture, the RL learning structure makes it possible to choose the best course of action. Honest extraction was formerly thought to be the most effective mining technique; however, it has now been shown that there are greater benefits to departing from fair mining. You may define the mining issue as a process of Markov decision making (MDP) and solve it to find the best mining technique. The values of many variables, which are hard to find and subject to change in the future, are necessary to solve the MDP. In order to continuously develop a mining approach with efficiency that is close to the ideal mining method, the research makes use of reinforcement learning. The extraction MDP problem's non-linear function objective is resolved using a novel multifunctional RL method. The multifunctional RL mining method may reach optimum performance across time-varying blockchain platforms, according to research results.

With the rapid expansion of consumer electronics and the Internet of Things (IoT) ecosystem, ensuring secure communication in resource-constrained edge devices has become a critical challenge, especially under the threat of quantum attacks. Although lattice-based cryptographic schemes provide post-quantum security, they often suffer from large key storage and high computational overhead, limiting their applicability in edge environments. To address these challenges, we propose an identity-based linkable ring signcryption scheme constructed on the post quantum Lattice cryptography, tailored for edge-enabled IoT systems. The scheme employs a trapdoor generation algorithm to establish the master keys, utilizes compact Gaussian sampling to derive private keys for ring members, and integrates rejection sampling to generate user signatures with strong anonymity. Furthermore, a key encapsulation mechanism ensures identity protection during encryption. Security analysis under the Random Oracle Model (ROM) demonstrates the scheme's resistance to forgery and its linkability property. Performance evaluation shows that the proposed approach achieves smaller public keys, reduced encryption/decryption costs, and enhanced efficiency, making it well-suited for lightweight IoT and consumer electronics applications in the post-quantum era.