Active Motor Noise Cancellation (AMNC) is a noise-reduction feature shipped in commercial fused deposition modeling (FDM) 3D printers. Because it suppresses the acoustic emissions that side-channel attacks exploit, it has security-relevant side effects, though we find no evidence it was designed as a security control. We present a duration-controlled evaluation of side-channel leakage on AMNC-equipped hardware, using a public dataset of 144 synchronized acoustic and vibration recordings from two Bambu Lab printers across 12 object classes. Spectral analysis confirms suppression is measurably active: the motor-resonance band rises only 4.92 dB above its background-relative baseline during printing. Leakage nonetheless survives it. Acoustic classification is at chance for 30-second windows (11.11

Accurate annual average daily traffic (AADT) estimation underpins infrastructure planning, environmental assessment, and safety analysis; however, coverage gaps persist in areas with scarce continuous counters. We propose a data-fusion framework that imputes missing AADT values by integrating crash records, clean energy infrastructure metadata, and roadway features. We evaluate a diverse set of machine learning families for tabular regression, including tree ensembles (random forest, gradient boosting), linear and regularized regression baselines, k-nearest neighbors, and shallow neural networks, and compare them against a stacked ensemble that learns to combine base-model predictions.Models are trained using an AutoML framework to standardize preprocessing, validation, and ensembling. Across traffic-only, energy-only, and hybrid feature sets, the stacked ensemble consistently achieves the lowest prediction error and remains robust across low-, medium-, and high-volume traffic regimes. On the combined feature set, the model achieves a mean absolute error (MAE) of approximately 570 vehicles/day, a root mean squared error (RMSE) of approximately 1415 vehicles/day, a mean absolute percentage error (MAPE) value of 3.73, and an R2 value close to 0.99 under cross-validation evaluation. These results demonstrate that principled ensembling and multi-source data fusion substantially improve AADT imputation performance, particularly in data-limited settings such as uncounted or sparsely monitored roadways.

Advanced Driver Assistance Systems (ADAS) enhance road safety by supporting drivers through warnings and control assistance; however, their effectiveness depends on accurate and interpretable recognition of driver behavior. This study proposes an ensemble learning framework that integrates deep learning and Explainable Artificial Intelligence (XAI) to classify driver behavioral states using eye-tracking data. Eye-region images are processed using ResNet50, DenseNet201, and InceptionV3 for feature extraction, and the extracted features are fused using an XGBoost classifier. The proposed framework achieves an overall classification accuracy of 94.90% and an average AUC of 0.97 across multiple gaze-related driving behavior classes. DenseNet201 contributes strong discrimination of fixation-related patterns, ResNet50 provides robust and generalizable spatial representations, and InceptionV3 captures multi-scale features associated with subtle gaze deviations. The ensemble model leverages these complementary representations to improve robustness and reduce misclassification. SHAP-based analysis revealed that upper and lateral gaze features positively contribute to attentive driving states, while lower-gaze regions, blink-related features, and pupil-related representations are the dominant contributors to distraction-related behaviors. These findings provide interpretable insight into how eye-tracking features drive model decisions. By combining quantitative performance gains with feature-level explanations, the proposed framework enables transparent, behavior-aware driver state monitoring and supports the development of adaptive and interpretable ADAS capable of informed intervention and control handover.

Autonomous vehicles (AVs) promise transformative safety and mobility benefits, yet their deployment and operation in urban environments remain constrained by persistent safety concerns, especially at signalized intersections with complex pedestrian interactions. Right-turn maneuvers are particularly challenging due to occlusion-induced visibility limitations caused by urban infrastructure elements such as buildings, vegetation, and parked vehicles, which restrict the AV’s surrounding surveillance and hinder timely pedestrian detection. While vehicle-based sensors like LiDAR and cameras provide foundational awareness, they fall short under occluded conditions. Vehicle-to-Everything (V2X) communication, encompassing Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Pedestrian (V2P), offers a cooperative perception solution, allowing AVs to access critical information beyond their direct sensor field. This paper presents a MATLAB-based simulation framework designed to evaluate AV-pedestrian interactions at signalized intersections under diverse V2X communication and visibility conditions, replicating a real-world intersection in Jersey City, NJ. The framework enables real-time data sharing between AVs, infrastructure, and pedestrians to simulate detection performance and warning message exchange during right-turn scenarios. Results indicated that V2P enables early alerts with signal strength between 80%–100% allowing the ego-vehicle to decelerate safely to 5 m/s. V2I and V2V supported early detection through Road-Side-Unit (RSU) and other vehicles, speed reduction, and ensured timely response through shared warning messages. While the framework does not propose new algorithms, its primary contribution lies in the integrated simulation design that simultaneously evaluates all three V2X communication types. These findings provide actionable insights into enhancing transportation safety policy to improve AV pedestrian safety.

This paper presents a deep learning framework for classifying traffic signs as ‘safe’ or ‘not safe’ based on retroreflectivity performance relative to Manual on Uniform Traffic Control Devices (MUTCD) federal safety thresholds. A conditional Generative Adversarial Network (cGAN) overcomes dataset scarcity and class imbalance (86.9% safe vs 13.1% unsafe), achieving a Frechet´ Inception Distance score of 11.36 and enabling ablation-verified synthetic-only training equivalent to real-only training (98.41% vs 98.44%). A Convolutional Neural Network (CNN) trained on synthetic data achieves 91.52% clean accuracy; adversarial training restores accuracy to 70.98% under the most damaging compound attack combining the Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) (FGSM+PGD). Two Hybrid Neural Networks—HNN1 (classical-quantum) and HNN2 (quantum-classical)—achieve 95.98% clean accuracy, with HNN1 showing a statistically significant improvement over CNN (p = 0.040). Geographic validation on two independent datasets—510 samples from the Tiny Laboratory for Intelligent and SafeAutomobiles (LISA) dataset (United States) and 2,670 samples from the German Traffic Sign Recognition Benchmark (GTSRB) (Germany)—reveals that HNN2 retains 70.75% accuracy on GTSRB versus 49.21% for CNN; however, this advantage does not replicate on LISA, where all architectures converge to 44–53%, indicating that generalization is dataset-dependent rather than universal. These findings validate adversarial training effectiveness and reveal both the potential and current boundary conditions of quantum-classical approaches for future transportation systems.

Road fatalities pose significant public safety and health challenges worldwide, with pedestrians being particularly vulnerable in vehicle-pedestrian crashes due to disparities in physical and performance characteristics. This study employs explainable artificial intelligence (XAI) to identify key factors contributing to pedestrian fatalities across the five U.S. states with the highest crash rates (2018-2022). It compares them to the five states with the lowest fatality rates. Using data from the Fatality Analysis Reporting System (FARS), the study applies machine learning techniques - including Decision Trees, Gradient Boosting Trees, Random Forests, and XGBoost - to predict contributing factors to pedestrian fatalities. To address data imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) is utilized, while SHapley Additive Explanations (SHAP) values enhance model interpretability. The results indicate that age, alcohol and drug use, location, and environmental conditions are significant predictors of pedestrian fatalities. The XGBoost model outperformed others, achieving a balanced accuracy of 98%, accuracy of 90%, precision of 92%, recall of 90%, and an F1 score of 91%. Findings reveal that pedestrian fatalities are more common in mid-block locations and areas with poor visibility, with older adults and substance-impaired individuals at higher risk. These insights can inform policymakers and urban planners in implementing targeted safety measures, such as improved lighting, enhanced pedestrian infrastructure, and stricter traffic law enforcement, to reduce fatalities and improve public safety.

Driver fatigue causes up to 20% of traffic fatalities. However, EEG-based detection systems remain limited by physiologically weak behavioral labels, poor cross-subject generalization, and insufficient interpretability. We present the EEG-based Fatigue Analysis and Detection Engine (EEG-FADE) to address these challenges. EEG-FADE introduces Ratio-Based Thresholding (RBT), deriving labels from theta/beta ratio dynamics calibrated to a subject's baseline alert-state EEG. This is validated via sensitivity analysis (Cohen's kappa > 0.84) and temporal independence testing (mean time-label correlation r = 0.044). A 480-dimensional feature set (spectral, temporal, complexity) is extracted and evaluated using three pipelines: AutoML (AutoGluon), Bidirectional LSTM, and Graph Neural Networks with Squeeze-and-Excitation attention (GNN-SE). Experiments on the Cao et al. dataset (23 subjects, 98,848 segments) reveal a performance ranking reversal: AutoML achieves the highest global stratified F1 (0.908), whereas GNN-SE achieves a significantly higher leave-one-subject-out (LOSO) F1 (0.792; p < 0.001) with the lowest inter-subject variance (SD = 0.086). Ablation of 150 TBRrelated features confirms non-TBR features sustain LOSO F1 scores of 0.65-0.69, proving classification avoids label circularity. Cross-architecture interpretability analysis (permutation importance, integrated gradients, SE attention) converges on frontal channels (Fz, F4, Fp1) as dominant fatigue markers, aligning with known neurophysiology. EEG-FADE uniquely addresses physiological labeling, cross-subject generalization, and interpretability, recommending the GNN-SE pipeline for diverse-user deployment without individualized calibration.

The integration of Internet of Things (IoT) devices into space information networks introduces unprecedented security challenges, necessitating advanced methods for information systems risk assessments. Some proposed approaches in terrestrial networks, predominantly based on Euclidean distance metrics, often fall short in capturing the nuanced and multidimensional nature of risks and vulnerabilities in the unique context of space environments. This paper proposes an integrative risk management methodology which leverages both Euclidean distance metrics and probabilistic Monte Carlo-based models for evaluating the likelihood/frequency and impact/severity of vulnerabilities within space IoT systems. Leveraging vulnerability data from established sources such as the NIST and VARIoT databases, our approach simulates the use of the methodology in a device scenario, allowing for a deterministic and probabilistic assessment of vulnerability criticality. The deterministic component provides a practitioner-oriented heuristic security posture index, while the probabilistic component introduces a formally specified stochastic framework grounded in empirical vulnerability data. The key contributions of this work lie in combining deterministic and distance-based methods with a stochastic framework that accounts for the uncertainty and complexity of space-based IoT networks. By incorporating probabilistic simulations, our model provides more objective and data-informed criticality ratings, supporting decision-making processes to secure space information systems. The importance of this research is underscored by its potential to structure and simplify risk management in space IoT systems, providing an accessible and data-informed framework suited to the evolving threat landscape for novice risk analysts. This work fills an identified gap in space IoT risk assessment practice and sets the stage for future developments in secure space IoT deployments.

LiDAR-based perception in autonomous systems is fundamentally limited by sparse vertical sampling and further degraded by structured beam dropout caused by occlusions, sensor faults, or reduced-cost LiDAR hardware. These degradations disrupt vertical geometric continuity and negatively affect downstream perception tasks such as object detection, localization, and scene understanding. Existing reconstruction approaches often struggle to balance reconstruction accuracy with the computational efficiency required for real-time autonomous operation. This paper presents SuperiorGAT, a graph attention-based framework for reconstructing missing elevation information in sparse LiDAR point clouds under structured beam loss. The proposed method models LiDAR scans as beam-aware graphs and enhances standard graph attention networks using gated residual fusion and lightweight feed-forward refinement to improve vertical reconstruction fidelity without increasing network depth. The effectiveness of SuperiorGAT is evaluated on multiple KITTI environments, including Person, Road, Campus, and City, as well as through cross-dataset validation on nuScenes with lower vertical resolution. Additional experiments under severe structured sparsity further evaluate robustness in a 16-beam-equivalent sensing condition. Results demonstrate that SuperiorGAT achieves lower overall reconstruction error and improved geometric consistency compared to interpolation-based methods, PointNet-based models, and standard GAT baselines while maintaining computational efficiency suitable for real-time perception pipelines.

Cyber-physical systems rely on sensors, communication, and computing, all powered by integrated circuits (ICs). These ICs are vulnerable to malicious hardware attacks, with hardware Trojans being one of the stealthiest threats. Trojans are malicious implants in the circuitry, which are often inserted during design or fabrication stages. This stealthy addition remains dormant until triggered and might cause functional disruptions or sensitive information leakage once triggered. Traditional IC validation methods, such as functional testing and logic analysis, usually fail to capture these subtle anomalies because hardware Trojans are intentionally designed to mimic normal circuit behavior. They often remain dormant under standard test vectors, and the changes they bring into power consumption, timing, and area are minimal. We present a dual-domain feature extraction strategy that combines time-domain features with frequency-domain characteristics of power traces. For the detection, we created an artificial intelligence (AI) based robust Trojan detection framework that integrates traditional machine learning models, such as random forest (RF), gradient boosting (GB), naïve bayes (NB), and deep learning models, such as deep neural network (DNN), long short-term memory (LSTM), and graph neural network (GNN). In this study, we consider these models as baseline AI models to detect Trojan-infected circuits via side-channel power analysis. We employed a stacked ensemble classifier that integrates the distinct strengths of the six baseline models used in this study. After evaluating our stacking ensemble-based detection method on the advanced encryption standard (AES)-Trojan benchmark, which covers diverse Trojan types, the results demonstrate that the ensemble method consistently outperformed all six baseline models. The ensemble-based detection method achieved a macro-averaged area under the receiver operating characteristic (ROC) curve (AUC) of 0.987, while remaining golden-chip-free, meaning it does not rely on a trusted reference IC for baseline comparison. Instead it detects anomalies directly from observable characteristics of untrusted chips, such as side-channel emissions.

Roads and bridges form the backbone of transportation infrastructure, requiring accurate vehicle-weight information for safe, cost-effective design, maintenance, and management. Traditional Weigh-In-Motion (WIM) systems provide high-quality weight data but remain costly to install and maintain, leaving large portions of the U.S. highway network without direct weight coverage. To address this limitation, this study develops and compares six predictive models for estimating Gross Vehicle Weight (GVW) of heavy-duty vehicles (FHWA Classes 4–13) using 2021 WIM datasets from three geographically distinct sites in New York (Site 01), California (Site 02), and Texas (Site 03). The modeling framework includes baseline approaches (Lookup Table and Multiple Linear Regression), statistical models (Class-Specific Regression and Generalized Additive Model), and ensemble methods (Random Forest and Extreme Gradient Boosting). Input features were restricted to vehicle class and number of axles, identified as the dominant predictors through correlation analysis and SHapley Additive exPlanations (SHAP)-based feature selection. Model performance was evaluated using R2, RMSE, MAE, MAPE, and the Transferability Index (TI) under repeated random-split, supplementary time-based, and repeated cross-site validation frameworks, with results summarized using mean values, standard deviations, and 95% confidence intervals. The results demonstrate a consistent performance hierarchy across all study sites, with ensemble models substantially outperforming the baseline and statistical approaches. XGBoost achieved the highest in-site predictive accuracy (R2 up to 0.84), whereas Random Forest consistently demonstrated superior cross-site generalization (TI up to 0.95). These findings demonstrate that accurate and transferable GVW estimation can be achieved using only two universally available vehicle attributes, providing a scalable, interpretable, and computationally efficient framework for freight monitoring, pavement-load assessment, and infrastructure management on non-instrumented road segments.

Deep Neural Networks (DNNs) have demonstrated remarkable success across a wide range of tasks, particularly in fields such as image classification. However, DNNs are highly susceptible to adversarial attacks, where subtle perturbations are introduced to input images, leading to erroneous model outputs. In today's digital era, ensuring the security and integrity of images processed by DNNs is of critical importance. One of the most prominent adversarial attack methods is the Fast Gradient Sign Method (FGSM), which perturbs images in the direction of the loss gradient to deceive the model. This paper presents a novel approach for detecting and filtering FGSM adversarial attacks in image processing tasks. Our proposed method evaluates 10,000 images, each subjected to five different levels of perturbation, characterized by ϵ values of 0.01, 0.02, 0.05, 0.1, and 0.2. These perturbations are applied in the direction of the loss gradient. We demonstrate that our approach effectively filters adversarially perturbed images, mitigating the impact of FGSM attacks. The method is implemented in Python, and the source code is publicly available on GitHub for reproducibility and further research.

Security-focused program testing typically focuses on crash detection and code coverage while overlooking additional system behaviors that can impact program confidentiality and availability. To address this gap, we propose a statistical framework that combines embedding-based anomaly detection, resource usage metrics, and resource-state distance measures to systematically profile software behaviors beyond traditional coverage-based methods. Leveraging over 5 million labeled samples from 50 Python programs, we evaluate how these independent scoring terms distinguish among different sources of input, including Large Language Model (LLM)-generated inputs, and demonstrate how standard statistical tests (e.g., Kolmogorov-Smirnov and Kendall's tau) confirm their effectiveness. Our findings show that LLM-generated samples can trigger diverse behaviors but are often less effective at exploring resource usage dynamics (CPU, memory) compared with conventional fuzzing. However, combining LLM outputs with existing techniques broadens behavior coverage and reveals commonalities between commercial LLM outputs. We provide open-source tools for this evaluation framework, demonstrating the potential to refine software testing by integrating behavior metrics into security-testing workflows.

Deep learning (DL)-based image classification models are essential for autonomous vehicle (AV) perception modules since incorrect categorization might have severe repercussions. Adversarial attacks are widely studied cyberattacks that can lead DL models to predict inaccurate output, such as incorrectly classified traffic signs by the perception module of an autonomous vehicle. In this study, we create and compare hybrid classical-quantum deep learning (HCQ-DL) models with classical deep learning (C-DL) models to demonstrate robustness against adversarial attacks for perception modules. Before feeding them into the quantum system, we used transfer learning models, alexnet and vgg-16, as feature extractors. We tested over 1000 quantum circuits in our HCQ-DL models for projected gradient descent (PGD), fast gradient sign attack (FGSA), and gradient attack (GA), which are three well-known untargeted adversarial approaches. We evaluated the performance of all models during adversarial attacks and no-attack scenarios. Our HCQ-DL models maintain accuracy above 95% during a no-attack scenario and above 91% for GA and FGSA attacks, which is higher than C-DL models. During the PGD attack, our alexnet-based HCQ-DL model maintained an accuracy of 85% compared to C-DL models that achieved accuracies below 21%. Our results highlight that the HCQ-DL models provide improved accuracy for traffic sign classification under adversarial settings compared to their classical counterparts.

Traffic emissions significantly impact near-road air quality and public health. This research applies a Bayesian modeling framework to investigate these impacts using high-resolution traffic and air pollutant data from an urban corridor in Columbia, South Carolina. Despite a data collection period truncated by the COVID-19 lockdown, the Bayesian approach successfully identified significant predictors and quantified model uncertainty. Employing Bayesian Model Selection and Averaging enhanced prediction accuracy and evaluated model uncertainty. Findings indicate that higher temperatures and increased moisture levels elevate particulate matter (PM1.0, PM2.5, PM10) concentrations, while traffic speed significantly affects nitrogen dioxide (NO2) levels. Specifically, higher average traffic speeds (indicative of smoother flow) correspond to lower NO2 concentrations, suggesting that less congested conditions reduce NO2 emissions. This study highlights the robustness of Bayesian methods for generating reliable air quality insights even under data-constrained conditions. The findings underscore the importance of traffic flow management (e.g., reducing congestion) for mitigating near-road NO2 exposure and provide a basis for developing targeted public health strategies.

Large trucks substantially contribute to work zone-related crashes, primarily due to their large size and blind spots. When approaching a work zone, large trucks often need to merge into an adjacent lane because of lane closures caused by construction activities. This study aims to enhance the safety of large truck merging maneuvers in work zones by evaluating the risk associated with merging conflicts and establishing a decision-making strategy for merging based on this risk assessment. To predict the risk of large trucks merging into a mixed traffic stream within a work zone, a Long Short-Term Memory (LSTM) neural network is employed. For a large truck intending to merge, it is critical that the immediate downstream vehicle in the target lane maintains a minimum safe gap to facilitate a safe merging process. Once a conflict-free merging opportunity is predicted, large trucks are instructed to merge in response to the lane closure. Our LSTM-based conflict prediction method is compared against baseline approaches, which include probabilistic risk-based merging, 50th percentile gap-based merging, and 85th percentile gap-based merging strategies. The results demonstrate that our method yields a lower conflict risk, as indicated by reduced Time Exposed Time-to-Collision (TET) and Time Integrated Time-to-Collision (TIT) values relative to the baseline models. Furthermore, the findings indicate that large trucks that use our method can perform early merging while still in motion, as opposed to coming to a complete stop at the end of the current lane prior to closure, which is commonly observed with the baseline approaches.

With urbanization and rising vehicle numbers, road safety has become increasingly critical. Robust, trajectory-level risk assessment is essential for next-generation active safety systems, accident prevention, autonomous driving, and intelligent transportation networks. This paper presents a novel framework for driver behavior classification using Topological Data Analysis (TDA) — a mathematical approach for analyzing high-dimensional data — via persistent homology applied to vehicle trajectory data. Traditional methods often struggle with the complexity of such data, but TDA captures topological features that reveal subtle, meaningful behavioral patterns. Using the HighD dataset, we train a class-weighted XGBoost classifier on persistence image (PI) features, achieving 96.8% overall accuracy, macro-F1 = 0.93, and retaining 87% F1 on the minority Aggressive class. Unsupervised K-means clustering of the same PI features naturally separates the data into three behavioral clusters whose ANOVA-verified risk profiles align with the MOR-defined classes, confirming the behavioral relevance of the topological descriptors. These results provide empirical evidence that PI features capture safety-critical structure more effectively than raw kinematics and demonstrate the robustness and scalability of TDA for analyzing large, noisy datasets. The proposed approach shows strong potential for real-time driver monitoring, risk assessment, and data-driven transportation management, with implications for traffic safety, autonomous systems, and personalized insurance.

Pedestrian safety is a critical public health priority, with pedestrian fatalities accounting for 18 prevalence of distracted walking, exacerbated by mobile device use, poses significant risks at signalized intersections. This study utilized an immersive virtual reality (VR) environment to simulate real-world traffic scenarios and assess pedestrian behavior under three conditions: undistracted crossing, crossing while using a mobile device, and crossing with Light-emitting diode (LED) safety interventions. Analysis using ANOVA models identified speed and mobile-focused eye-tracking as significant predictors of crossing duration, revealing how distractions impair situational awareness and response times. While LED measures reduced delays, their limited effectiveness highlights the need for integrated strategies addressing both behavioral and physical factors. This study showcases VRs potential to analyze complex pedestrian behaviors, offering actionable insights for urban planners and policymakers aiming to enhance pedestrian safety.

In transportation planning, Zero-Vehicle Households (ZVHs) are often treated as a uniform group with limited mobility options and assumed to rely heavily on walking or public transit. However, such assumptions overlook the diverse travel strategies ZVHs employ in response to varying trip needs and sociodemographic factors. This study addresses this gap by applying a weighted Latent Class Cluster Analysis (LCCA) to data from the 2022 National Household Travel Survey (NHTS) to uncover distinct mobility patterns within the ZVH population. Using travel mode and trip purpose as indicators and demographic, economic, and built environment variables as covariates, we identified three latent classes :Shared mobility errand workers (36.3

The Federal Highway Administration (FHWA) mandates that state Departments of Transportation (DOTs) collect reliable Annual Average Daily Traffic (AADT) data. However, many U.S. DOTs struggle to obtain accurate AADT, especially for unmonitored roads. While continuous count (CC) stations offer accurate traffic volume data, their implementation is expensive and difficult to deploy widely, compelling agencies to rely on short-duration traffic counts. This study proposes a machine learning framework, the first to our knowledge, to identify optimal representative days for conducting short count (SC) data collection to improve AADT prediction accuracy. Using 2022 and 2023 traffic volume data from the state of Texas, we compare two scenarios: an 'optimal day' approach that iteratively selects the most informative days for AADT estimation and a 'no optimal day' baseline reflecting current practice by most DOTs. To align with Texas DOT's traffic monitoring program, continuous count data were utilized to simulate the 24 hour short counts. The actual field short counts were used to enhance feature engineering through using a leave-one-out (LOO) technique to generate unbiased representative daily traffic features across similar road segments. Our proposed methodology outperforms the baseline across the top five days, with the best day (Day 186) achieving lower errors (RMSE: 7,871.15, MAE: 3,645.09, MAPE: 11.95