We propose a novel dance generation technology called Dance Choreography Generation from Music with Dancer-specific Style (SDCGM) that automatically generates new dance choreography to match a specified music track and dancer-specific style. This technology is achieved through a transformer-based conditional diffusion model framework, where both music and style information are utilized as input conditions. The results of evaluation experiments indicate that compared to existing state-of-the-art models, the dances generated by the SDCGM model exhibit higher quality in terms of naturalness and are more capable of reflecting the specified dancer-specific style appropriately. Furthermore, user evaluations of a prototype application suggest that the dance generation functionality provided by SDCGM could be beneficial for professional dancers in terms of mastering dances and supporting choreography creation. By utilizing SDCGM, anyone can specify their favorite music and dancer-specific style to generate a wide variety of attractive dances.

Distributed file systems employ two deduplication methods: intra-node deduplication and inter-node deduplication. Inter-node deduplication, which eliminates redundant data across multiple nodes, achieves a superior capacity reduction compared to intra-node deduplication, which is confined to a single node. However, inter-node deduplication has the risk to store duplicate data on a different node from the original file, which increases the number of inter-node communications and reduces the data read performance. To mitigate this challenge, we propose an approach that retains duplicate data within the original file until storage capacity constraints necessitate eviction, thereby minimizing inter-node communication. The evaluation results show that the proposed method improves I/O performance by 3.2 times compared to the conventional method in the training workload of AI image processing. Furthermore, our evaluation results indicate that the release of the cache in units of duplicate chunks may result in a decline in data read performance when compared to the conventional method. However, this issue can be addressed by releasing the cache in units of files.

We have proposed a new communication technology, Brain Model Hyper Communication Mechanism (BMHCM). This technology models the information processing mechanism of the brain, especially the visual mechanism, and divides it into sending and receiving sides for communication. A feature of this method is that it sends the input without processing it, enabling more advanced processing, and then processes it hierarchically on the receiving side. The Modified Brain Model Hyper Communication Mechanism (MBMHCM) adds a feedback loop and mechanism for direct transmission to the upper layer; thus, skipping the hierarchy. This study formulates MBMHCM and implements a MBMHCM-based three-layer concentration maintenance system.

Hand drawing in immersive virtual environments works best on a virtual surface aligned with a counterpart physical surface. However, physical surfaces are rarely portable and therefore hard-to-align with the virtual surfaces in the air. To overcome this limitation, we propose a mechanical pen device that changes the length to render encountered haptic feedback. We evaluated our proof-of-concept device in a user study and published the 3D printable data for reproducibility. We designed a trajectory tracing task on 0-/15-/30-degree slanted canvases and compared drawing with and without our encountered haptic rendering in terms of tracing accuracy, usability, and workload. The results revealed that our method has the potential to afford better line tracing performance and reduce workload. We discuss how precise device length controls can contribute to improving usability and decreasing workload on a steep canvas.

Smullyan introduced many combinators, including the combinators M, O, and P. Recently, we showed that P is not terminating by a tree automaton using our SAT-based implementation. From observations on the results, we found similarities in the form of the M, O, and P rules, as well as similarities of the terms that disprove their termination. In addition, Curry and Feys introduced a family of combinators S_n. In this paper, we disprove the termination of a family of combinators that, generalizes the combinators M, O, and P, by a family of tree automata. Furthermore, we disprove the termination of the family of the combinators S_n by family of tree automata.

This paper proposes a new method to improve the characteristics of the triangulated irregular network (TIN) map deformation. Specifically, the similarity transformation-based interpolation method is adopted to improve the characteristics of the smoothness of the point-position transformation function from the original map to the deformed map. The experimental results show that the proposed method actually realizes the smoothness on the edges of the triangles in the TIN. However, it is also indicated that unsmoothness still remains only on the vertices of the triangles.

Recently, the development of the Metaverse has increased the demand for immersive head-mounted displays (immersive HMD). Using immersive HMD in daily life requires user awareness of the surrounding people and objects while ensuring a high sense of immersion and visibility. This study proposes HidEye, an interaction method that enables users to easily switch between virtual and real-space content by covering one eye. Furthermore, applications are also developed in which covering one eye is used to change the field of view in a virtual space. The performance of the system is also verified through performance evaluations and user studies.

Conventional in-vehicle information systems for rear-seat passengers have been tailored to tourist areas with POIs, proving effective in those contexts. However, these are ill-suited for daily use in urban environments, where vehicles rapidly pass numerous POIs. In such settings, conventional systems face two primary challenges: (1) displaying vast amounts of information on limited screen space, and (2) allowing users to select desired information quickly as the vehicle moves at high speeds. To overcome these challenges, we developed a digital car window system that reimagines the external scenery as “video content.” This system integrates playback control, letting users pause, rewind, or fast-forward the exterior view video. It also features automatic selection of the nearest POI and a steady-paced ticker display of items, up to one item every two seconds. We evaluated the system through laboratory and on-road experiments. In the lab, the system achieved a mean SUS score of 70.5, exceeding the benchmark of 68. On-road testing revealed that repeated use enhanced user comfort and enabled intuitive information searches. Compared to smartphone-based approaches, our system significantly improved the ease of taking photographs, the ease of retrieving information, the enjoyment of riding, and the level of interest in the surroundings. This research advances the practical deployment of in-vehicle digital displays in urban environments.

This study proposes an air-handwriting recognition technique that enables seamless, touchless text input in environments such as public displays, in-vehicle systems, and AR glasses. By leveraging intuitive handwriting, the approach expands the design space for human-computer interaction in contexts where physical keyboards or touchscreens are impractical. It offers potential privacy advantages by eliminating the need for voice or touch input, and promotes inclusivity by supporting arbitrary vocabulary without relying on predefined lexicons or prior training. The proposed system, Mobile-AeroText, employs a single-stage object detection network based on GELAN. It transforms fingertip trajectories into binary images while simultaneously detecting and recognizing character regions, enabling robust word-level recognition without explicit boundary gestures. In an evaluation with 25 participants and 1, 600 words, Mobile-AeroText achieved a word recognition rate of 91.44%, a character recognition rate of 95.86%, and an average latency of 417 milliseconds on a CPU. Subjective assessments yielded a System Usability Scale score of 78.5 and a NASA-TLX score of 33.6, indicating high usability and low cognitive load. Overall, this study presents a practical “write-anywhere” input method that addresses a fundamental challenge in human-computer interaction: enabling expressive, accessible, and low-burden text input in scenarios where traditional methods are unavailable or limited.

The rapid growth of educational video content has increased demands on storage systems in universities. Although MinIO offers a cost-effective object storage solution, its default configurations might not provide optimal upload performance. This study investigates optimization strategies for MinIO file upload performance under small-start educational environments, while examining the impact of chunk sizes and parallel upload numbers across different network conditions (including high latency variations and packet loss). Through systematic experiments, we found that optimal configurations vary by file size: 100MB files perform best with 8MB chunks and 16-32 parallel uploads, 1GB files achieve optimal performance with 64MB chunks and 16-32 parallel uploads, while 10GB files perform best with 128MB chunks and 16-32 parallel uploads. These settings showed an average 55% improvement in upload throughput under normal conditions, but a 400% improvement was demonstrated in high-latency environments, providing practical guidelines for educational institutions deploying object storage systems.

In-store advertising, such as digital signage and posters, is a crucial method that influences customer behavior. While effectiveness is often evaluated by displaying ads on a store-by-store basis and comparing outcomes for those exposed to ads and those not, obtaining individual ad exposure data is costly, making it difficult to conduct causal inference with individual-level treatment variables. A common approach to address this issue is to perform causal inference considering non-compliance, treating visitors to stores with advertising as the treatment group and similar customers who visited comparable stores as the control. In this setting, a popular estimator is the ratio of two Difference-in-Differences (DID) estimates: one for the outcome and one for the treatment variable. However, prior studies assumed the DID estimate for the treatment variable is known, which is not always true. To address this, we propose a causal inference method using the fact that, for binary treatment, the DID estimate of the treatment variable represents the change in the proportion of compliers in the treatment group. Our method uses a Gaussian Mixture Model to estimate this proportion. This approach allows estimation of the treatment effect on compliers even when individual ad exposure data is unobserved.

Simultaneous Localization and Mapping (SLAM) is a technique that estimates a self-position while simultaneously constructing a map of the surrounding environment. In SLAM, errors in the estimated self-position can significantly affect overall performance. Global Bundle Adjustment (GBA) is a widely used optimization process intended to correct such errors and improve estimation accuracy. While GBA often enhances trajectory accuracy, in some cases it can introduce substantial degradation. To address this problem, we propose Motion Check (MC), a method that compares the estimated trajectories pre- and post-GBA. When a significant discrepancy in trajectory shape is observed, the method judges that GBA has caused a deterioration in accuracy. By identifying points where accuracy degrades due to GBA and applying appropriate corrective actions, Motion Check effectively suppresses such deterioration. In evaluations using the EuRoC dataset, Motion Check demonstrated its effectiveness by reducing the maximum trajectory error by up to 81.52cm and the root mean square error by up to 3.40cm.

Pencil-and-paper puzzles have been capturing the interest of the scientific community in recent years, especially the ones invented and popularized by the Japanese publisher Nikoli. The reason for this interest lies in the inherent difficulty of solving these puzzles, which has led to many studies on their computational complexity. Various Nikoli puzzles have been shown to be NP-complete, and a consistent decreasing trend has been observed before for the amount of time between the publication of a new Nikoli puzzle and the establishment of the puzzle's first hardness result (from gaps of almost 20 years for most of the company's older puzzles to gaps of less than 5 years for most of the newer ones). In this paper, we prove that Nagenawa, a Nikoli puzzle, is NP-complete.

Image style transfer, a prominent application of generative AI, has significantly impacted digital content creation. However, the proliferation of style transfer models makes it challenging for users to search and find a model that generates a desired style. Usually, searching for an appropriate model is a manually intensive and computationally prohibitive task. To address this, we propose an effective method for style transfer model retrieval. Our method takes an image exhibiting a target style as a query and ranks a collection of available models based on their ability to reproduce that style. Our approach uses a Vision Transformer to encode the query image's features into patch-level embeddings. Concurrently, style transfer models are represented as learnable embedding vectors. A transformer encoder then fuses query image and model embeddings to compute a relevance score, indicating the model's suitability for generating style similar to the query image's. Our method is optimized end-to-end using a hybrid loss function that combines Binary Cross Entropy (BCE) with a Learning-to-Rank objective. To evaluate the proposed method, we constructed a benchmark dataset of 10,000 images, generated from 100 unique style transfer models applied to 100 distinct content images. Our experiments demonstrate the effectiveness of the proposed method's retrieval performance as well as its efficiency.

Although recent supercomputers have been improving their computational performance, achieving performance scaling with respect to the number of nodes is not easy due to long inter-node communication latency. Many attempts have been made to hide communication latency and maintain strong scalability even for dense matrix multiplication. Matrix multiplication is an ideal candidate for benchmarking the performance of supercomputers. The Cerebras CS-2 system is an accelerator for deep learning with the world's largest chip, the wafer-scale engine 2 (WSE-2). The WSE-2 can be considered a distributed memory system that comes with 745, 500 processing elements connected in a low-latency 2-D mesh topology. This paper presents the effective maximum performance, weak and strong scaling performance, and proposes a performance model for single-precision matrix multiplication on the CS-2. We observed the maximum performance of 349.0TFlops/s (matrix size: 33,000×33,000, used PEs: 750×750), performance per watt of 79.66GFlops/W, and a weak scaling efficiency of 1.00. The mean absolute percentage error between our performance model and the actual measurement was 9.2%.

Transaction processing is widely used in various applications, including credit card payments and social data analysis. To deal with these applications, scalable concurrency control protocols have been proposed. Modern techniques mainly target SEARCH and UPDATE operations and YCSB and TPC-C benchmarks are used for their evaluation. Unfortunately, the issues of improving the performance of SCAN, which is mandatory for large-scale data analysis, and avoiding phantom anomalies have not been addressed in such techniques. In this paper, we propose a novel concurrency control protocol for the efficient execution of transactional wide-range SCAN based on the optimistic concurrency control protocol. The proposed protocol, Node-locking, acquires locks of the leaf nodes in the tree index during SCAN to prevent phantoms in the read phase. The proposed protocol avoids deadlocks due to locking by applying the WaitDie protocol to index nodes. On a workload containing SCAN transactions reading 10,000 records, Silo with the proposed protocol shows up to 743 times performance improvement over the vanilla Silo protocol.

Black hole accretion disks serve as the central engines driving high-energy phenomena such as X-ray binaries, active galactic nuclei, and gamma-ray bursts, and give rise to complex physical processes such as turbulence induced by magnetorotational instability, intense heating, and the formation of relativistic jets. Traditional one-dimensional theoretical models and general relativistic magnetohydrodynamic (GR-MHD) simulations have successfully captured many aspects of these phenomena, but they fall short in reproducing the effects of radiative cooling and radiation pressure, which are critical in high accretion rate environments. To overcome these limitations, general relativistic radiation magnetohydrodynamics (GR-RMHD) simulations have been developed to incorporate the interplay between radiation and plasma dynamics. However, the inclusion of radiative transfer significantly increases the computational complexity, requiring advanced computational strategies for high-fidelity multidimensional simulations. In this study, we present a GR-RMHD code implemented using CUDA Fortran and MPI that achieves up to 39-fold speedup over conventional CPU implementations, with nearly linear scalability as the number of GPUs increases. This acceleration enables detailed analysis of the complex dynamics of black hole accretion disks and jet formation, providing new insights into the energy production mechanisms in high-energy astrophysical systems. We discuss the evolution of theoretical models and numerical simulations from GR-MHD to GR-RMHD, describe the algorithms and GPU implementation of our approach, and provide quantitative performance evaluations.

This study proposes YuShi, a two-stage method that rapidly generates synthetic radar rainfall data for regions with no historical experience of torrential rain, thereby enabling preevent testing and training. In the first stage, a terrain- and meteorology-conditioned log-Gaussian Cox process probabilistically synthesizes the rainfall-core occurrence field. In the second stage, a diffusion-based UNet augmented with low-rank adaptation (LoRA) is used to create the contour mask and the intensity texture separately. A threefold validation using ten Japanese heavy-rain events (2018-2024) revealed that the hit rate of the generated cores was 1.00 within a 25km tolerance, the wind direction error was less than 0.05rad, and the nationwide spatial error (Wasserstein_area) was reduced to approximately 54% of the observed value, confirming high fidelity in direction and large-scale distribution. The remaining errors are concentrated along coastal regions, where high-intensity cells are overly smoothed and regional biases in the outer radius appear. These issues are expected to be alleviated by region-specific scaling of the occurrence intensity and by introducing terrain indices and wind convergence channels in the texture stage. These results provide synthetic rainfall fields that can serve as foundational inputs for future cyber-physical evaluations of the Disaster IT Testbed.

Phishing attacks have become increasingly sophisticated, necessitating effective methods for real-time detection. This study examines phishing site characteristics, focusing on SSL certificate validity periods and user authentication forms. Based on a dataset of 1,047 unique phishing URLs collected from multiple sources, we found that 89.8% of the phishing sites had certificate validity periods of 90 days or less, with Let's Encrypt (53.9%) and Google Trust Services LLC (26.5%) being the most prevalent certificate issuers. To effectively prevent this threat, we developed a Proof-of-Concept (PoC) browser extension that leverages certificates and user authentication forms required for information theft as key indicators for real-time phishing detection. The PoC extension achieved a detection rate of 94.5%. Comparative analysis demonstrated that our approach outperformed several existing solutions, including Google Safe Browsing, in phishing site detection. While the proposed method demonstrates high effectiveness, we also discuss its limitations. This study demonstrates the potential of approaches that use certificates and user authentication forms for real-time detection, particularly for phishing sites targeting financial institutions and e-commerce platforms, offering a simple yet effective countermeasure.

Pedestrian navigation systems are essential not only for exploring unfamiliar cities but also for discovering unexpected experiences in daily life. While modern systems visualize point-to-point routes enriched with contextual information, such as building icons, terrain shapes, and nearby points of interest, deploying and evaluating them through in-the-wild experiments remains challenging. In this paper, we aim to discuss how researchers develop in-the-wild and customizable navigation systems and the challenges they face when conducting in-the-wild experiments. Based on our field study conducted in Hibiya Park using a system that combines volunteered geographic information and standard routing engines, this paper reports the two key findings. First, publicly available geographic data often lack the resolution; thus, we need to carefully setup navigation environments. Second, although precise tracking is technically possible, current positioning devices still face the inaccurate positioning results; thus, we need to moderate navigation systems to be realized in real-world environments, particularly when implementing original mobility experiences via navigation. Based on these findings, we argue for domain-specific refinement of open-source geographic data to design immersive mobility experiences.

Text-to-image (T2I) models have garnered attention for their expressive capabilities, highlighting the need for robust evaluation metrics. Text-image alignment, a critical evaluation aspect, has recently received broad attention, leading to various evaluation methods and benchmarks. Among T2I models, stable diffusion has become widely used for generating images from arbitrary prompts; however, their performance varies based on prompt difficulty, which prior studies have not comprehensively addressed. To bridge this gap, we propose a template-based evaluation method for controllable difficulty (TEM-CD). Our approach defines prompt difficulty levels and integrates attention maps with a vision-language model to enable precise evaluation across diverse image classes, attributes, and styles. We validated the method through two experiments. First, a user study confirmed that its results aligned more closely with human assessments than conventional metrics. Second, we applied it to three stable diffusion models (SD1.4, SD1.5, and SD2). Results showed that SD2 achieved superior expressiveness, particularly in attribute generation. Despite overall similarities, our method identified differences in generative capability. These findings highlight the robustness and reliability of the proposed approach, demonstrating its effectiveness as an evaluation method for stable diffusion models.

Microwave power transfer (MPT) has emerged as a method of wireless power transmission for Internet of Things (IoT) devices. A rectifying antenna can convert microwaves into power for the devices while simultaneously facilitating data communication. While most research has focused on efficient power transfer for a single IoT device, there have been few studies considering scenarios involving multiple IoT devices. This paper proposes an efficient and fair MPT system for multiple IoT devices: more delivered power at the same transmit budget; more even per-device satisfaction with worst-user protection. The proposed system achieves both efficiency and fairness by mitigating power attenuation for each IoT device and optimizing power distribution based on the varying demands of IoT devices. Our results demonstrated that the proposed system outperformed an existing system in terms of both efficiency and fairness for each IoT device.

Malware security appliances are critical defenses against the growing number of malware attacks organizations face. However, malware authors have integrated sandbox evasion techniques into modern malware, enabling attacks that may bypass security appliances and products. Furthermore, adversaries can design Customized Malware that only reveals its malicious payload when the identifier of the target-specific system can be verified and bypass sandboxes and even non-targeted hosts. This approach presents another burden and challenge for malware sandbox engineers, as security has not kept pace with these developments. This paper presents a targeted attack scenario and evaluates the risk of this threat, while providing ideas for countermeasures. We introduce the idea that attackers can leverage publicly available personally identifiable information (PII) from the target system as identifiers, particularly targeting hosts that store business email addresses on their PCs. To first evaluate Applicability, we investigated a set of desktop applications and specified 18 popular applications that store email addresses in their files or directories. To evaluate Identifiability, we implemented a survey tool to access these applications and record whether email addresses were found. The survey was conducted with 218 MTurk workers. Among 143 workers using either of the applications, email addresses could be found from all of their PCs. Additionally, a test with 9 laboratory members and staff confirmed the ability to identify target-specific email address from 16 desktop applications. In addition to email addresses, usernames used on social media platforms are effective in identifying target systems. Therefore, by preparing several social networking service applications installed on a test environment, we specified 3 popular applications that store usernames in their files or directories. Finally, to evaluate Stealthiness, we implemented dummy malware samples that search for the target host's email address or usernames in the executing environment and prevents unpacking the malicious payload if the mark is absent. We show that two commercial malware security appliances and a malware analysis solution would fail to unpack the malicious payload. To mitigate this attack scenario, we discuss countermeasures from both the sandbox and user perspective, including enhanced sandbox alerts for search behavior and inserting dummy email addresses and/or usernames into target systems. We also contacted relevant security appliance vendors to inform them of the potential threat and provided proof-of-concept programs to assist in their preparation.

This paper introduces the development of a web-based educational word-guessing game aimed at enhancing English language learning through interactive gameplay. Leveraging large language models (LLMs) to generate contextual hints, semantic scores, and word lists, the game fosters an engaging and dynamic learning environment. Building upon a previous metaverse-based implementation, this web approach improves accessibility, scalability, and deployment ease. The system architecture incorporates modern web technologies like React, Node.js, and Tailwind CSS to ensure a seamless user experience, while addressing key challenges such as LLM integration, real-time feedback, and progressive hinting mechanisms. Preliminary results suggest improved usability and broader adoption potential across diverse educational settings. This paper contributes to the field by demonstrating how web-based platforms, combined with LLMs, can offer scalable, effective, and gamified learning experiences, highlighting the potential for further advancements in language education technology.

This study examines the relationship between cultural activities in daily life and the mental and physical well-being of respondents. We conducted a survey on cultural activities and health among residents of Takayama City, Gifu Prefecture, Japan. The survey includes both structured responses and free-text descriptions covering demographic data, hobbies, artistic activities, and emotional experiences. In particular, we focus on emotionally moving experiences, referred to in Japanese as 感動 (kandõ), which denotes experiences that strongly impress and deeply move one emotionally. Given the combination of numerical and textual data, direct statistical analysis is challenging. To address this, we convert the survey responses into text format and develop a regression model using BERT, a large-scale language model, to estimate mental and physical well-being. Training data are categorized into four groups: demographic data, hobbies, artistic activities, and emotional experiences. Our findings show that the regression model based on emotional experiences provides the most accurate estimations of both mental and physical well-being. SHAP analysis further reveals that experiences related to nature, such as “flowers, ” “nature, ” “growth, ” and “sky, ” are strongly associated with well-being, reflecting the significance of culturally and emotionally experiences (感動) in everyday life.

This paper presents the design and implementation of a frontend system that dynamically provisions user-specific execution environments on a container-based virtualization infrastructure. The system responds to SSH connection requests by automatically deploying the corresponding environment using Docker or Kubernetes. We extended ContainerSSH, open-source software for container management. This extension enables users to select their desired environment from multiple administrator-predefined options via a single SSH endpoint. The implemented enhancements incorporate database-driven environment management and SSH public key authentication to ensure security and flexibility. We evaluated the system's performance under concurrent connections of up to 100 users in an AWS cloud environment. The results demonstrated that users could log in within 5 seconds at 20 concurrent connections. Distributing connections across multiple nodes via load balancing achieved practical responsiveness. Our additional features introduced only 4 percent overhead compared to the baseline system. Analysis of internal processing revealed that container operations dominated total latency, suggesting optimization through pre-warming container pools during peak usage.

With the rapid advancement of information technology these days, there is an increasing demand to cultivate human resources capable of identifying and solving issues effectively through information technology, further underscoring the importance of information education. In this context, “learning by quiz creation, ” which learners create quizzes on their own, is expected to be highly effective in fostering the application of knowledge and skills. While there have been many studies on methods to support learning by quiz creation, this study proposes a new approach to support learning by quiz creation using generative artificial intelligence (GenAI). We designed quiz creation processes (QCPs) for three typical types of quiz creation (story-, quiz-, and solution-based) aimed at deepening learners' understanding. Furthermore, we developed quiz creation prompts to enable GenAI to guide these QCPs effectively while also supporting learners' proactive quiz creation. To verify the effectiveness of the proposed method, we set two research questions: (1) Can GenAI adequately guide the proposed QCPs? and (2) Can GenAI with prompts enhance learning effectiveness of quiz creation? We conducted an experimental application involving university students, and the results suggest that GenAI can serve as a suitable guide for learners in the initial stages of each QCP, with particular effectiveness observed in solution-based quiz creation. Additionally, GenAI equipped with quiz creation prompts received higher evaluations from learners compared to general GenAI, demonstrating its potential to enhance the learning experience.

Identifying the physical properties of objects is a crucial technique in several industrial, security, and inspection applications. In this study, we propose a novel method for non-invasively estimating multiple properties of a target object, including scale, shape, and density, by leveraging channel state information (CSI) obtained from a commercial-off-the-shelf (COTS) Wi-Fi device. Our proposed method employs a multi-branch convolutional neural network (CNN) architecture that learns the intricate correspondence between ambient signal propagation and the target properties. The novelty of our method lies in its recursive mechanism: each network branch outputs the classification results for a specific property, which then serves as input to other branches. This multi-branch and recursive strategy allows for a comprehensive understanding of the target, as the estimation of each property is refined by mutually referencing the results of the others. We evaluate our proposed method using a tabletop testbed with COTS Wi-Fi devices and 3D printed objects. We compare our method with vision-based and CSI-based baselines, and demonstrate that our method achieves a better classification accuracy of 99% for all object properties under various illumination conditions and the distance between the object and the Wi-Fi devices.

Reducing traffic accidents and congestion remains a major challenge, especially at intersections which often cause traffic bottlenecks. Connected and Automated Vehicles (CAVs) are expected to improve traffic safety and efficiency through cooperative driving based on communication with other vehicles and infrastructure. While various cooperative driving methods for unsignalized intersections have been proposed, most of these methods assume that all vehicles in the system are CAVs. However, it is also necessary to consider mixed traffic environments involving both CAVs and human-driven vehicles since full CAV market penetration is not expected for several decades. There are several challenges in considering human-driven vehicles such as speed deviation and acceleration specifications. In this paper, we propose a cooperative driving method in mixed traffic at unsignalized intersections based on Virtual Platoon. We determine minimum following distances based on vehicle types and human-driven vehicles' target speed which is indicated to the drivers. Simulation results using a traffic simulator demonstrate that the proposed method successfully avoids collisions in mixed traffic and reduces average travel time by up to 18.7 seconds compared to signal control.

Secure messaging (SM) protocols allow users to communicate securely over an untrusted infrastructure. Alwen et al. formally defined the key agreement protocol used in secure group messaging (SGM) as continuous group key agreement (CGKA) at CRYPTO 2020. In their CGKA protocol, all of the group members have the same rights to execute algorithms, and a trusted third party is needed. On the other hand, some applications such as a broadcasting service may have an administrator in the group. When the administrator as the group manager (GM) is distinguished from other group members, i.e., in a one-to-many setting, it would be better for the GM and the other group members to have different authorities. We propose a new CGKA protocol, called CGKA with flexible authorization (CGKA-FA) by incorporating a ratcheting digital signature scheme (Cremers et al. at USENIX Security 2021) into the existing CGKA protocol and demonstrate that such a simple modification allows us to provide flexible authorization. Our proposed protocol does not need an external public key infrastructure (PKI) functionality as a trusted third party. So no third party outside the group handles confidential information of users and thus no one except group members can read messages regardless of key updates. In our CGKA-FA protocol, it is possible to set privileges of users in a group flexibly, which is highly compatible with the field of broadcasting.

Trajectories recorded via the Global Positioning System (GPS) can be used to estimate the transport mode, such as train, bus, or car. This study estimated the modes from sparsely sampled data designed to resemble real-world trajectories. Different sampling frequencies, preprocessing methods, and other classification techniques were compared as sensitivity analyses within the sparse dataset. The proposed model achieved an accuracy of 74.8% and an F1-score of 75.2% at a 5-minute sampling frequency. Speed and geographic information contributed the most to the classification task. The collection of additional trajectory data may further improve accuracy and influence the outcomes of comparative analyses.