In the high-stakes arena of aerial combat-a domain defined by extreme dynamics and unforgiving physical constraints-UAV swarms are currently squeezed between two extremes: the "tactical short-sightedness" of Multi-Agent Reinforcement Learning (MARL) and the "inference lag" of Large Language Models (LLMs). While MARL struggles to internalize the complex maneuverability priors required for expert flight, LLMs are simply too heavy to meet millisecond-level control demands. We bridge this gap by introducing a cognitive synergetic hierarchical framework that decouples strategic reasoning from tactical execution. Our architecture splits the workload between a "Strategic Brain" and a "Tactical Torso." For the Brain, we utilize a synergy between DeepSeek-R1 (70B) and its 7B distilled counterpart to create a collaborative inference engine. By capitalizing on the inherent sparsity of tactical logic in air combat, we implemented a speculative decoding mechanism that achieves an effective boost in decision throughput while maintaining the deep logic of the full 70B model. For the Torso, we developed an enhanced MAPPO algorithm that processes relative pose graphs via graph attention. By integrating a KL-divergence constraint into the loss function, we essentially force agents with different payloads-like scouts and attackers-to evolve specialized tactical personalities within a shared latent space. Experimental results using the JSBSim high-fidelity 6-DOF engine demonstrate that the swarm does more than just improve its exchange ratio. Further t-SNE manifold analysis and Chain-of-Thought visualizations confirm that our architecture successfully aligns symbolic intent with raw physical control. Most notably, through our "decision-reflection-evolution" loop, the system proved it could diagnose its own failures, and iteratively refine its own tactical instructions.
Headache attributed to anxiety or depressive disorders was added to the appendix of the International Classification of Headache Disorders, 3rd edition (ICHD-3) in 2018. However, its specific clinical features have not yet been systematically characterized, limiting its early recognition and optimal clinical management. We conducted an observational study between March 2024 and June 2025 at a headache center in China. Following stringent exclusion of primary headache disorders and other secondary headache causes, 101 patients meeting diagnostic criteria for headache attributed to anxiety and depressive disorders (HA-ADD) were enrolled. Demographic data, detailed headache characteristics, and psychiatric symptoms were collected through structured face-to-face interviews. The cohort was predominantly female (70.3%; n = 71) with a median age of 36 years. The characteristic HA-ADD phenotype was defined by: (1) pain characteristics: bilateral pain (72.3%) with dull quality (92.1%) and moderate-to-severe intensity (53.5% moderate; 42.6% severe); (2) location: temporal (47.5%) and parietal regions (35.6%); (3) temporal pattern: 60.4% of attacks lasted less than 4 h; (4) associated symptoms: phonophobia (79.2%), restlessness/agitation (58.4%), dizziness (57.4%), and nausea (51.5%); (5) trigger factors: emotional fluctuations (71.3%) and poor sleep quality (57.4%) were predominant. Notably, 73.2% of patients experienced severe headache-related functional impairment (HIT-6 score >60). Headache attributed to anxiety and depressive disorders presents a distinct clinical profile characterized by bilateral, dull, moderate-to-severe pain with short attack duration and prominent associated symptoms including phonophobia and restlessness. This phenotypic signature differentiates HA-ADD from common primary headache disorders and provides clinical markers that may facilitate earlier diagnosis and appropriate therapeutic intervention.
Chronic tension-type headache (CTTH) is a prevalent primary headache disorder that significantly affects quality of life. Hyperacusis, an abnormal intolerance to everyday sounds, is frequently observed in patients with CTTH, suggesting potential shared neural mechanisms. However, the clinical characteristics of patients with CTTH and comorbid hyperacusis remain poorly characterized. This study aimed to investigate the clinical characteristics of CTTH patients with comorbid hyperacusis, focusing on headache profiles, auditory sensitivity patterns, and psychological correlates, and to identify factors independently associated with hyperacusis in this population. A cross-sectional study was conducted at the outpatient department of the Third Affiliated Hospital of Qiqihar Medical College from January 2022 to December 2024. We consecutively enrolled 234 patients with CTTH diagnosed according to International Classification of Headache Disorders, 3rd edition (ICHD-3) criteria were consecutively enrolled. Hyperacusis was confirmed via otolaryngological evaluation and audiometric testing (pure-tone thresholds and loudness discomfort levels [LDL]). Demographic data, headache features (frequency, intensity, and duration), hyperacusis distress severity, and psychological measures (Generalized Anxiety Disorder-7 [GAD-7], Self-rating Somatic Symptom Scale [SSS], and Pittsburgh Sleep Quality Index [PSQI]) were analyzed. Univariate analyses (Student's t-test and Chi-squared test) and multivariate logistic regression adjusted for potential confounders were performed to compare patients with (n = 47) and without (n = 187) hyperacusis. Hyperacusis prevalence was 20.09% (47/234). Compared to non-hyperacusis patients, hyperacusis patients exhibited longer headache duration (13.9 ± 3.15 versus 8.11 ± 2.74 years, P = 0.01), higher weekly headache frequency (7.30 ± 1.99 versus 4.19 ± 1.2, P = 0.005), more pericranial tenderness (95.74% versus 77.54%, P = 0.008), and poorer sleep quality (PSQI: 14.01 ± 1.13 versus 9.17 ± 0.51, P = 0.004). After adjusting for age, sex, headache duration, and pain intensity in multivariate logistic regression, higher headache frequency (odds ratio [OR] = 1.42, 95% confidence interval [CI]: 1.12-1.79, P = 0.003) and higher PSQI scores (odds ratios = 1.38, 95% CI: 1.15-1.66, P = 0.001) remained independently associated with hyperacusis. Common triggering sounds included children crying (61.7%) and traffic noise (42.55%), with predominant emotional responses being irritability (76.6%) and anxiety (65.96%). LDL classification (analyzed as an ordinal variable) was correlated with headache duration (P < 0.05), while hyperacusis distress severity was linked to headache frequency and PSQI scores (P < 0.05). CTTH with hyperacusis represents a distinct subgroup characterized by a history of prolonged headaches, increased attack frequency, and significant sleep disturbances. The independent associations identified suggest that sound sensitivity is closely linked to headache burden and sleep quality; however, causal relationships require longitudinal investigation. Integrated management addressing both headache and auditory hypersensitivity is warranted.
Plants and animals respond to pathogen attack by mounting innate immune responses that require intracellular nucleotide-binding leucine-rich repeat (NLR) proteins. These immune receptors detect pathogen infection by sensing virulence effector proteins. However, how receptors evolve new recognition specificities remains poorly understood. We found that the plant NLR MLA3 (Mildew locus a 3) has evolved to recognize a pathogen effector by acting as a molecular mimic of an effector virulence target, thereby triggering an immune response. By introducing the mimic's binding interface into the wheat stem rust resistance protein SR50, we bioengineered a chimeric receptor with dual recognition activities that conferred resistance to two major cereal pathogens in barley transgenic lines. These results demonstrate that molecular mimicry by immune receptors can be harnessed to engineer multiple disease resistance.
Hereditary transthyretin-related (ATTRv) amyloidosis may involve the central nervous system (CNS) years after liver transplantation, causing transient focal neurological episodes (TFNEs). Their mechanisms and electrophysiological correlates remain unclear. We retrospectively analyzed 155 patients with ATTRV30M (p.V50M) amyloidosis patients who underwent liver transplantation and had at least one electroencephalogram (EEG). TFNEs were categorized into five clinical subtypes. EEGs from outpatient and hospital settings were included. TFNEs occurred in 76.8% of patients, most commonly as seizure-like (37.0%), transient ischemic attack-like (25.2%) or migraine-like (24.4%) episodes. Stroke-like TFNEs and bilateral tonic-clonic seizures were rarely the initial presentation (13.4%). Anti-seizure medications were prescribed to 77.3% of patients with TFNEs and 26.9% required two or more drugs. Focal slow EEG activity was detected in 40.0% of patients and was associated with TFNEs (OR = 2.91) and reduced survival (HR = 2.18). During acute TFNEs, focal slowing predominated, whereas epileptiform patterns were infrequent (19.5%). Cognitive impairment correlated with posterior dominant rhythm slowing and focal slow activity, particularly in language and executive domains. TFNEs are common after liver transplantation in ATTRv amyloidosis. Focal slow EEG activity may indicate advanced CNS amyloid deposition, cognitive impairment and poorer prognosis.
Background and objectives Hepatitis A virus (HAV) remains a major cause of jaundice outbreaks globally, with 158.9 million infections reported in 2019 despite a 63% decline in mortality since 1990. Improved sanitation in many regions has shifted the age of primary infection from early childhood to older children and adults, increasing the risk of symptomatic and severe disease. In April-May 2025, we investigated an outbreak of HAV caused hepatitis that occurred in Rajur village, Ahilyanagar district, Maharashtra. A total of 327 suspected cases of acute hepatitis were reported, mainly affecting older children and young adults. Methods Suspected acute hepatitis cases were identified by clinical presentation and their epidemiological and clinical data were collected by questionnaire. Serum, stool, and potable water samples were collected and tested by anti-HAV IgM ELISA and/or real-time RT-PCR to confirm outbreak aetiology. Full HAV genome sequences were obtained from selected RT-PCR positive specimens. Results The age-specific attack rate for the 327 acute hepatitis cases was the highest in children aged 10-19 years (11.97%) and lowest in those ≥ 50 years. Anti-HAV IgM antibodies were detected in 32 of 45 (71.1%) available serum samples, and HAV RNA was detected in 7 of 22 (31.8%) serum and 19 of 31 (61.3%) stool samples tested. All 7 potable water samples tested negative for HAV RNA. Genomic analysis revealed strain details relevant to disease severity, providing insights into circulating HAV lineages in India. Interpretation and conclusions The findings highlight the need for continuous surveillance, genomic monitoring, and targeted prevention strategies to protect vulnerable, previously unexposed populations.
Migraine headaches are common and potentially disabling disorders, with several interventions available for prevention and symptom reduction. We explored the comparative effectiveness and tolerability of pharmacological prophylaxis for migraine through a network meta-analysis of randomised trials (RCTs). Our study design was a systematic review and network meta-analysis (PROSPERO registration CRD42023456915). We included randomised controlled trials of prophylactic pharmacological interventions that enrolled adults diagnosed with chronic and/or episodic migraine headaches. Medline, Embase, Cochrane Central, PsycINFO, Web of Science and Scopus from inception to 15 January 2026. Risk of bias was assessed using the modified Cochrane risk-of-bias tool 2.0 and the certainty evidence was evaluated by using the Grading of Recommendations Assessment, Development and Evaluation approach. We performed a frequentist network meta-analysis using a random-effects model to compare the efficacy of interventions. We included 199 RCTs (47 420 participants). Overall, 29 trials (14.6%) were at low risk of bias; an adequate random allocation sequence generation was reported in 92 trials (46.2%), and missing outcome data was the most common limitation (110 trials, 55.3%). Compared with placebo, calcium channel blockers (mean difference (MD) -1.78 (95% CI -2.96 to -0.60), moderate certainty), calcitonin gene-related peptide (CGRP)-targeted therapies (MD -1.69 (95% CI -2.16 to -1.23), high certainty) and beta-blockers (MD -1.50 (95% CI -2.54 to -0.47), moderate certainty) were the most effective in reducing monthly migraine days. Moderate certainty evidence suggests beta-blockers (MD -1.31 (95% CI -1.76 to -0.85)), calcium channel blockers (MD -1.11 (95% CI -1.65 to -0.57)), anticonvulsants (MD -1.12 (95% CI -1.66 to -0.58)) and CGRP-targeted therapies (MD -0.76 (95% CI -1.49 to -0.02)) probably reduce monthly migraine attacks. However, moderate to high certainty evidence found that patients were more likely to discontinue calcium channel blockers (relative risk (RR) 1.40, 95% CI 1.04 to 1.88) and anticonvulsants (RR 1.14, 95% CI 1.01 to 1.29), compared with placebo. When restricted to moderate or high certainty evidence, beta-blockers and CGRP-targeted therapies probably reduce migraine frequency and may be well-tolerated prophylactic options for migraine. Calcium channel blockers and anticonvulsants may also be effective for reducing migraine frequency but are less well tolerated by some patients. CRD42023456915.
In this paper, a distributed event-based control architecture is proposed to improve the security and environmental performance of cyber-physical smart grids. The strategy concurrently responds to coordinated cyberattacks, such as false data injection (FDI) and denial-of-service (DoS), and incorporates carbon emission trading (CET) into the optimization of the energy dispatch. It constructs a novel event-based resilient consensus algorithm (ERCA), which incorporates attack detection and recovery schemes into a distributed decision-making framework. The algorithm employs a trust-node-based correction strategy and reliable acknowledgment signaling to maintain reliable state estimation and coordination under communication interruptions and data falsification. By incorporating carbon-pricing directly into the local cost functions, the framework enables generation units and responsive loads to achieve economically efficient and low-carbon operation without centralized supervision. Convergence of the proposed method is rigorously established under simultaneous FDI and DoS attacks. Simulation studies on an IEEE 41-bus system confirm that the framework maintains power balance, stabilizes electricity prices, ensures consistency in reported emissions, and reduces overall carbon output, even in the presence of stealthy and disruptive cyber intrusions.
The proliferation of Internet of Medical Things (IoMT) devices in 6G-enabled healthcare networks introduces critical cybersecurity challenges, including expanded attack surfaces, device heterogeneity, and real-time security requirements that traditional perimeter-based frameworks cannot adequately address. This paper proposes a Blockchain-Enabled Zero-Trust Architecture (B-ZTA), guided by the MITRE D3FEND defensive ontology, integrating Zero Trust micro-segmentation, a Random Forest-based intrusion detection system, and a lightweight Proof-of-Authority blockchain for deterministic policy enforcement. The framework is validated through MATLAB-based simulations rather than physical IoMT deployment, across 30 independent trials under diverse cyberattack scenarios in a simulated 62-device hospital environment. Under the evaluated scenarios, results indicate a 99.10% Threat Neutralization Rate, mean enforcement latency of 76.68 ms, and 95th percentile latency of 249.86 ms, satisfying 6G eMBB and mMTC latency requirements for IoMT monitoring use cases. Quantitative benchmarking against five state-of-the-art frameworks yields a composite security score of 91.8. These findings suggest promise for the B-ZTA approach within the simulated environment, with physical testbed validation and adversarial robustness evaluation identified as priority directions for future work.
Mobbing in academic environments significantly affects individuals' professional and psychological well-being. Due to their position in the educational hierarchy, graduate students may be more vulnerable to mobbing. This study aimed to examine academic mobbing experiences among graduate nursing students and to explore the associated challenges and coping processes. This study used a mixed-methods design. Quantitative data were analysed using descriptive statistics, independent samples t-tests, one-way ANOVA, and Pearson correlation analysis. A total of 209 graduate nursing students participated in the quantitative component of the study. Twenty graduate students were included in the qualitative research. Qualitative data were obtained through semi-structured interviews and analysed using Colaizzi's descriptive phenomenological method. Quantitative analyses revealed that doctoral students and research assistants had significantly higher total mobbing scores (p < 0.05), while female students reported significantly higher scores only in attacks on self-presentation and communication (p < 0.05). Qualitative findings indicated that communication barriers and professional pressures were prominent facilitators of mobbing. Qualitative analyses revealed that mobbing was primarily perpetrated by supervisors, administrators, and academic staff. In particular, interruptions, professional devaluation, excessive workload, and pressure via digital platforms were commonly reported. These findings highlight the role of academic hierarchy and tenure type in shaping mobbing experiences. The findings indicate that doctoral students and research assistants reported higher levels of academic mobbing, while female students were more vulnerable specifically to communication-based mobbing behaviors. Quantitative findings were complemented by qualitative data, which provided contextual insights into how participants perceived and reported mobbing experiences. However, this integration remains primarily interpretative and does not imply causal relationships between the identified factors and mobbing experiences. Specifically, mobbing was primarily experienced through communication barriers, professional pressures, and psychological coercion. These findings suggest consistency between quantitative and qualitative results and contribute to a better understanding of academic mobbing. not applicable.
Acute intermittent porphyria is a rare metabolic disorder that often presents with recurrent abdominal pain and nonspecific gastrointestinal symptoms, frequently leading to diagnostic delays in emergency settings. We report the case of a 24-year-old woman with glucose-6-phosphate dehydrogenase deficiency diagnosed at birth and pituitary prolactinoma treated with cabergoline 0.5 mg twice weekly, initiated one month before symptom onset, who presented with recurrent severe epigastric pain and vomiting over the course of one year. Repeated laboratory investigations and contrast-enhanced computed tomography of the abdomen and pelvis were unremarkable, and symptoms required opioid analgesia during multiple emergency department visits. A subsequent metabolic evaluation revealed elevated urinary porphyrin precursors, confirming acute intermittent porphyria. Following diagnosis, the cabergoline dose was reduced to 0.25 mg twice weekly with close monitoring of prolactin levels. During two years of follow-up, she experienced two recurrent attacks that were successfully managed with 2 L of 10% dextrose administered intravenously over three hours, along with opioid analgesia. This case highlights the importance of considering acute intermittent porphyria in patients with recurrent unexplained abdominal pain and emphasizes the possible role of medication-related and endocrine factors as indirect triggers, particularly in resource-limited settings where definitive therapy may not be available.
Familial Mediterranean Fever (FMF) is highly prevalent in Mediterranean countries and represents a substantial proportion of pediatric rheumatology patients. Colchicine, the mainstay of treatment, requires lifelong daily use, yet long-term adherence remains challenging, leading to recurrent attacks and increased healthcare burden. This study aimed to develop a freely downloadable mobile application to improve medication adherence by providing reminders and enabling direct communication with physicians. The application was designed as a structured medication-timing and monitoring tool, providing time-specific medication reminders, real-time intake confirmation ("taken/missed"), and digital adherence tracking. The user interface was tailored to the needs of patients and parents, considering children's attention span, parental workload, and health literacy. Following pilot testing in seven volunteer patients, the application was refined based on user feedback. Physicians used a web-based panel to register patients, define prescriptions and treatment schedules, and assign calendar-based appointments. Medication adherence was assessed using the MASIF scale, and disease activity was evaluated using the AIDAI score. A total of 93 patients with FMF were included in the study. Before application use, adherence was low, with a median MASIF score of 44 and most patients demonstrating poor adherence. After implementation, MASIF scores significantly improved at 3, 6, and 12 months, with a marked reduction in poorly adherent patients (p < 0.001). AIDAI scores and attack frequency significantly decreased, and no patient had an AIDAI score >9 during follow-up. Agreement between application-measured adherence and MASIF scores was excellent (ICC >0.90). Mobile reminder applications are low-cost, accessible digital health tools with strong potential to enhance medication adherence in FMF patients.
Knowledge corruption attacks occur when an attacker injects a small amount of malicious text into the knowledge database of a retrieval-augmented generation (RAG) system, thereby inducing the Large Language Model (LLM) to generate attacker-chosen target answers for specific target questions. Traditional knowledge corruption attacks are susceptible to semantic deviation and interference due to their single-generation strategy, in which malicious text is directly injected into the RAG system after generating it, resulting in limited concealment. This paper proposes a novel fine-grained, hierarchical, multi-round iterative semantic optimization attack method (FHM-ISO) for RAG systems. This method can continuously optimize the generated text by selecting different prompts based on the semantic relationships between the original question and the generated content, thereby guiding RAG systems to generate effective malicious texts. Meanwhile, we introduce a hybrid similarity scaling mechanism in FHM-ISO that integrates non-linear activation and linear scaling to optimize the evaluation of semantic distances, thereby overcoming the limitations of conventional rigid metrics in optimization. Experimental results demonstrate that the proposed method significantly improves the effectiveness of attacks against RAG systems across multiple public datasets, achieving an average attack success rate of 68% when only one piece of malicious text is injected per question.
In the recently emerging field of nonabelian group-based cryptography, a prominently used one-way function is the conjugacy search problem (CSP), and two important classes of platform groups are polycyclic and matrix groups. In this paper, we discuss the complexity of the conjugacy search problem (CSP) in these two classes of platform groups. We produce a polynomial time solution for the CSP in a finite polycyclic group with two generators, and show that a restricted CSP is reducible to a discrete logarithm problem (DLP). In matrix groups over finite fields, we use the Jordan decomposition of a matrix to produce a polynomial time reduction of an A-restricted CSP, where A ⊆ G L n ( F q ) is a cyclic subgroup, to a set of O ( n 2 ) DLPs over an extension of F q . For polycyclic groups with two generators we show that the CSP where conjugators are restricted to a cyclic subgroup is either equivalent to a DLP in some ( Z / N Z ) * or to an exponential diophantine integer equation. Using our general results, we demonstrate concrete cryptanalysis algorithms for each of these three schemes. We believe that our methods and findings are likely to allow for several other heuristic attacks in the general case.
The pervasive co-occurrence of microplastics (MPs) and per- and polyfluoroalkyl substances (PFAS) poses a significant challenge for remediation technologies. Electrochemical advanced oxidation processes like electro-Fenton (EF) are promising for MPs degradation, yet their efficacy in complex, co-contaminant systems remains poorly understood. This work revealed a previously overlooked mechanism by which perfluorooctanoic acid (PFOA) and its derivatives severely inhibited the electrochemical degradation of polyethylene terephthalate (PET)-MPs in a pyrite-modified heterogeneous EF system. The inhibition evolved through two distinct phases: an initial phase (0-5 h) dominated by reactive oxygen species (ROS) and electron competition, followed by a subsequent phase (5-10 h) governed by interfacial shielding. This shielding arose from the formation of a persistent fluorine-rich layer on the MPs' surface, facilitated mainly by hydrogen bonding between the oxidized MPs' surface and PFOA-derived intermediates (i.e., short-chain fluorotelomer carboxylic acids), which blocked further ROS attack. Such inhibition led to a drastic reduction in the degradation efficiency of PET-MPs by > 50% at 50 mg/L PFOA and was still effective under more environmentally relevant conditions (1-10 μg/L PFOA). These findings underscore that in heterogeneous co-contaminant systems, interfacial interactions could induce a more profound and persistent inhibitory effect than homogeneous competition alone, providing critical insights for designing effective remediation strategies for complex environmental matrices.
Infectious bacteria remain among the most plausible agents for deliberate biological attack because they combine environmental robustness, low infectious dose, and the capacity for rapid, severe disease. This review synthesizes current knowledge on major bacterial biowarfare threats including Bacillus anthracis, Yersinia pestis, Francisella tularensis, Brucella spp., and Clostridium botulinum across three levels: transmission routes, molecular pathogenesis, and defense strategies. We summarize how aerosol, food-water, and vector-borne pathways, together with globalization and urban crowding, shape outbreak potential, and contrast non-contagious threats such as inhalational anthrax with highly transmissible pneumonic plague. At the mechanistic level, we highlight convergent virulence platforms capsules and stealth surfaces, intracellular survival programs, type III/VI secretion systems, and toxins such as anthrax lethal toxin and botulinum neurotoxin that delay immune recognition and compress the window for effective intervention. We then review advances in surveillance and countermeasures, including portable PCR and CRISPR-based diagnostics, next-generation anthrax vaccines, antibiotics and antitoxins for plague and botulism, and emerging decontamination technologies for persistent spores. Finally, we discuss the integration of these tools within CBRNE incident management and global biosecurity frameworks, emphasizing persistent gaps in environmental remediation, resistance surveillance, and capacity in low-resource settings. Together, these data define priorities for strengthening resilience to bacterial biowarfare and bioterrorism.
Accurate and scalable soybean crop health monitoring remains a major challenge in precision agriculture due to environment variability, inconsistent lighting conditions, and significant differences between the ground-level leaf imagery and UAV-based aerial imagery. Most existing deep learning approaches treat these two sensing modalities separately without properly exploring cross-scale feature transferability or measuring the domain gap that exists between the sensing scales. As a result, developing unified and deployment-ready crop health monitoring systems that can effectively leverage the more accessible leaf-level datasets, collected without specialized equipment or regulatory constraints, to improve UAV-scale inference remains difficult. In order to address this limitation, we propose a multi-scale soybean crop health assessment framework that integrates ground-level leaf imagery and UAV-based aerial imagery from the MH-SoyaHealthVision dataset across four health conditions, which include Healthy, Mosaic Virus, Pest attack, and Rust. CLAHE, Gray-World color constancy correction, and illumination normalization is incorporated into a structured pre-processing pipeline and further applied to reduce illumination bias and enhance cross-domain feature consistency. Six deep learning backbones were comprehensively evaluated for leaf-level classification, with MaxViT and ConvNeXt achieving the best performance. Their static weighted ensemble further improved accuracy to 87.08%. Cross-scale evaluation showed that zero-shot leap-to-UAV transfer achieved only 40% accuracy, thus highlighting the presence of a substantial domain shift. Fine-tuning improved UAV classification performance to about 97%, while a supervised contrastive learning framework specifically designed for cross-scale feature alignment further increased accuracy to approximately 98% with better convergence stability. Feature embedding analysis using PCA, t-SNE, and silhouette metrics demonstrated considerable improvements in inter-class separability (0.59 vs. 0.19) and reduced domain discrepancy (0.0336 vs. 0.114) under contrastive learning. These findings suggest that supervised alignment can generate more class-discriminative representations with lower cross-scale domain discrepancy, making them more suitable for scalable multi-scale cross-health monitoring.
In recent years, the internet of vehicles (IoV) has become an important enabler of intelligent transportation systems, providing vehicle-edge computing for latency-sensitive and computation-intensive vehicular applications. In such environments, the efficient offloading of tasks is pivotal; yet existing approaches primarily focus on optimising performance, often under the assumption of benign operating conditions or by employing static, trust-based mechanisms. However, these methods fall short in practical IoV implementations due to high mobility, short-lived connectivity, and the adversarial nature of IoV, where misbehaviour and resource exhaustion can substantially compromise the system's reliability and security. In response to these problems, we design SpiralEdge-IoV, a secure and adaptive task offloading framework that tightly integrates defence and optimisation. This framework embeds a logarithmic spiral defence (LSD) mechanism that models trust as a deepening, adaptive path over time, enabling online risk evaluation and incremental offensive action against suspicious parties. We integrate these risk scores into an in-built bio-inspired Addax-optimisation-based decision model (LSD-AddaxNet) to obtain security-aware multi-objective offloading decisions that not only minimise latency, energy consumption, and execution cost, but also maximise robustness. A combination of realistic vehicular edge-offloading traces and the VeReMi misbehaviour dataset is employed to conduct an extensive simulation-based evaluation, confirming the effectiveness of the proposed framework. Relative to representative optimisation- and learning-based baselines, SpiralEdge-IoV delivers up to 18% lower average task latency, reduces energy consumption by about 15%, and increases task success rates in adversarial settings by over 20%. In addition, the analyses on convergence and scalability demonstrate that the framework enables stable optimisation with acceptable runtime overhead in dense vehicular scenarios. SpiralEdge-IoV can be helpful for attack-resilient, low-latency IoV edge computing and is thus suitable for safety-critical vehicular applications and future intelligent transportation systems, as shown in the results.
To address the core issue of insufficient modeling of multi-time-scale temporal features in modern network attack detection, this study proposes a network attack detection model based on Clock-Work Recurrent Neural Network (CW-RNN). The model introduces a time-division processing mechanism, divides hidden layer neurons into modular structures with different clock frequencies, and integrates an attention mechanism to enhance feature extraction in key attack stages. An end-to-end intelligent detection architecture is constructed, which can adaptively capture characteristics of both short-term burst attacks and long-term latent attacks. Experimental verification is conducted on two benchmark datasets: University of New South Wales Network Benchmark 2015 (UNSW-NB15) and Canadian Institute for Cybersecurity-Intrusion Detection Systems 2018 (CSE-CIC-IDS2018). The results show that the detection accuracies of the proposed CW-RNN model on the two datasets reach 95.8% and 95.2%, respectively, and the macro-average F1-scores are 94.2% and 93.6%, respectively, which are significantly superior to mainstream benchmark models such as Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN). Meanwhile, the training time of the model is reduced by 25%, and the inference speed is increased by more than 18%, achieving dual optimization in detection accuracy and computational efficiency. Its modular multi-time-scale processing design provides an efficient and practical technical solution for network attack detection.