Deployment Environment Research Articles

Analysis of Key Factors Shaping the Performance and Reliability of AI Applications

Artificial Intelligence (AI) technologies have revolutionized a wide range of industries by providing cutting-edge solutions that streamline operations, improve decision-making accuracy, and deliver highly personalized user experiences. From automating routine tasks in manufacturing and logistics to enabling advanced data analysis in healthcare and finance, AI has become a critical tool for enhancing productivity and optimizing outcomes.This paper examines three critical AI applicationsautonomous driving, natural language processing (NLP), and facial recognitionto analyze the factors influencing their performance and reliability. The study identifies data quality, algorithm optimization, and deployment environment as pivotal elements that determine the effectiveness and fairness of these systems. Through a comparative analysis, this paper highlights how challenges such as data diversity, algorithmic bias, and environmental constraints impact system outcomes. It also explores strategies for improving accuracy, adaptability, and fairness in real-world settings. Given the rapid evolution of AI, the study emphasizes the importance of continuous innovation and incorporating user feedback into system design. Future research directions include analyzing the adaptive capabilities of AI systems and developing methods for better integrating user insights, ensuring AI's sustained advancement in addressing complex societal needs.

Enhancing Security in Software Design Patterns and Antipatterns: A Framework for LLM-Based Detection

The detection of security vulnerabilities in software design patterns and antipatterns is crucial for maintaining robust and maintainable systems, particularly in dynamic Continuous Integration/Continuous Deployment (CI/CD) environments. Traditional static analysis tools, while effective for identifying isolated issues, often lack contextual awareness, leading to missed vulnerabilities and high rates of false positives. This paper introduces a novel framework leveraging Large Language Models (LLMs) to detect and mitigate security risks in design patterns and antipatterns. By analyzing relationships and behavioral dynamics in code, LLMs provide a nuanced, context-aware approach to identifying issues such as unauthorized state changes, insecure communication, and improper data handling. The proposed framework integrates key security heuristics—such as the principles of least privilege and input validation—to enhance LLM performance. An evaluation of the framework demonstrates its potential to outperform traditional tools in terms of accuracy and efficiency, enabling the proactive detection and remediation of vulnerabilities in real time. This study contributes to the field of software engineering by offering an innovative methodology for securing software systems using LLMs, promoting both academic research and practical application in industry settings.

Self-trainable and adaptive sensor intelligence for selective data generation.

With the increasing integration of machine learning into IoT devices, managing energy consumption and data transmission has become a critical challenge. Many IoT applications depend on complex computations performed on server-side infrastructure, necessitating efficient methods to reduce unnecessary data transmission. One promising solution involves deploying compact machine learning models near sensors, enabling intelligent identification and transmission of only relevant data frames. However, existing near-sensor models lack adaptability, as they require extensive pre-training and are often rigidly configured prior to deployment. This paper proposes a novel framework that fuses online learning, active learning, and knowledge distillation to enable adaptive, resource-efficient near-sensor intelligence. Our approach allows near-sensor models to dynamically fine-tune their parameters post-deployment using online learning, eliminating the need for extensive pre-labeling and training. Through a sequential training and execution process, the framework achieves continuous adaptability without prior knowledge of the deployment environment. To enhance performance while preserving model efficiency, we integrate knowledge distillation, enabling the transfer of critical insights from a larger teacher model to a compact student model. Additionally, active learning reduces the required training data while maintaining competitive performance. We validated our framework on both benchmark data from the MS COCO dataset and in a simulated IoT environment. The results demonstrate significant improvements in energy efficiency and data transmission optimization, highlighting the practical applicability of our method in real-world IoT scenarios.

Design of Multimodal Obstacle Avoidance Algorithm Based on Deep Reinforcement Learning

The navigation obstacle avoidance method based on deep reinforcement learning has stronger adaptability and better performance compared to traditional algorithms in complex unknown dynamic environments, and has been widely developed and applied. However, when using multimodal information input, deep reinforcement learning strategy networks extract features that differ significantly between simulated and real world environments, resulting in poor algorithm output strategies and difficulty in transferring models obtained from simulation training to actual environments. To address the aforementioned issues, this article utilizes image segmentation to narrow the gap in environmental features, integrates multimodal information, and designs a deep reinforcement learning multimodal local obstacle avoidance algorithm, MMSEG-PPO, based on proximal strategy optimization algorithms. The algorithm is then ported to practical environments for deployment and testing. The experiment shows that the algorithm proposed in this article reduces the gap between the simulation environment and the actual environment, and has better performance and generalization when transplanted to the real world environment.

The Functional Connectome and Long-Term Symptom Presentation Associated With Mild Traumatic Brain Injury and Blast Exposure in Combat Veterans.

Mild traumatic brain injury (TBI) sustained in a deployment environment (deployment TBI) can be associated with increased severity of long-term symptom presentation, despite the general expectation of full recovery from a single mild TBI. The heterogeneity in the effects of deployment TBI on the brain can be difficult for a case-control design to capture. The functional connectome of the brain is an approach robust to heterogeneity that allows global measurement of effects using a common set of outcomes. The present study evaluates how differences in the functional connectome relate to remote symptom presentation following combat deployment and determines if deployment TBI, blast exposure, or post-traumatic stress disorder (PTSD) are associated with these neurological differences. Participants included 181 Iraq and Afghanistan combat-exposed Veterans, approximately 9.4 years since deployment. Structured clinical interviews provided diagnoses and characterizations of TBI, blast exposure, and PTSD. Self-report measures provided characterization of long-term symptoms (psychiatric, behavioral health, and quality of life). Resting-state magnetoencephalography was used to characterize the functional connectome of the brain individually for each participant. Linear regression identified factors contributing to symptom presentation including relevant covariates, connectome metrics, deployment TBI, blast exposure PTSD, and conditional relationships. Results identified unique contributions of aspects of the connectome to symptom presentation. Furthermore, several conditional relationships were identified, demonstrating that the connectome was related to outcomes in the presence of only deployment-related TBI (including blast-related TBI, primary blast TBI, and blast exposure). No conditional relationships were identified for PTSD; however, the main effect of PTSD on symptom presentation was significant for all models. These results demonstrate that the connectome captures aspects of brain function relevant to long-term symptom presentation, highlighting that deployment-related TBI influences symptom outcomes through a neurological pathway. These findings demonstrate that changes in the functional connectome associated with deployment-related TBI are relevant to symptom presentation over a decade past the injury event, providing a clear demonstration of a brain-based mechanism of influence.

Automated Virtual Machine Migration with Load Shifting Technique Avoiding Latency in Cloud Environment

Efficient management of Virtual Machines (VMs) in cloud computing is one key aspect to achieve the best utilization of resources and performance. The study is about improving Automated Virtual Machine Migration with a Load Shifting Technique in order to reduce latency through the cloud domain. This methodology dynamically distribute the load across multiple VMs by migrating their workloads to other physical hosts determined according certain resource utilization metrics available in real-time. The dynamic and unpredictable nature of cloud workloads requires a solution that is more responsive in the way it manages VMs. Live migration and pre-copy migration offer suboptimal performance in terms of time to migrate, downtime due to the prolonged copying period as heaps swell/collapse during live migrations. Utilizing sophisticated algorithms and real-time monitoring to avoid these pitfalls, our methodology allows us for proper load balancing with minimal latency. The primary elements of the proposed framework are a detailed monitoring module, an intelligent load shifting algorithm, high-performance migration method and feedback loop for continuous optimization. The monitoring module is responsible for collecting on-the-fly information about CPU, memory, network traffic and disk I/O such that can be exploited by the load shifting algorithm. The algorithm processes this data to decide which VMs should be migrated with the purpose of ensuring load balancing and decreasing latency achieved using these migrations. In order to test the proposed approach, comprehensive simulations and comparatives analyses utilizing other models were performed. This results in better latency reduction, increased resource utilization and improved overall system performance. In fact, the proposed method can handle not only current challenges in cloud computing such as load balancing and latency issued but also a scalable and effective approach for general cloud infrastructure usage. Finally, the Automated Virtual Machine Migration with Load Shifting Technique proves to be a reliable and successful solution for assisting in managing dynamic cloud workloads. Through real-time monitoring, advanced algorithms and continuous optimization this method has now been able to improve cloud services performance and reliability thereby allowing it to better serve service level agreements as well meet up with user satisfaction. We hope to adapt the algorithm and generalize it for even more advanced predictive models in future iterations, as well as evaluation of practical implementation on cloud-based deployment environments.

Migración de serverbox hacia un sistema de microservicios aplicando arquitectura vertical

Efficient data management is essential in public administration for the effective operation of municipal cadastral systems. In 2010, the Multifinality Cadastral Modernization Program was implemented, a partnership between the Puebla municipality government and Sistemas de Información Geográfica S.A. de C.V. (SIGSA), which included the ServerBox platform. Over time, the monolithic architecture of ServerBox showed limitations in maintenance, updates, and change deployment. This project hypothesizes that migrating ServerBox to a distributed microservices architecture, using a vertical approach, can overcome these challenges and improve operational areas of the system. The proposed design establishes a modular and scalable structure that facilitates system maintenance and evolution. The implementation includes the development of TagHelpers in ASP.NET Core to simplify HTML element creation and the use of snippets to streamline coding. Additionally, a JSON configuration file is defined to manage multiple deployment environments. This strategic approach to migrating ServerBox to a distributed and vertical architecture offers significant advantages in terms of scalability, maintenance, and software development efficiency, and can serve as a guide for similar projects in public administration.

Weighted Centroid Amorphous Algorithm for Position Error Minimization in WSN

ABSTRACTWireless Sensor Networks (WSNs) have numerous applications, one of which is localization. Localization is crucial for determining the position of unknown sensor nodes in the deployment environment. Localization techniques are broadly classified into two categories: range based and range‐free. Range‐free localization techniques are gaining popularity as they are easy to implement and do not require external hardware. In this proposed work, a hybrid localization algorithm named the Weighted Centroid Amorphous algorithm is proposed to reduce the position error in WSN. Instead of the conventional centroid algorithm, a weighted centroid algorithm is used. The weight in this work is considered as a function of the hop value and hop size estimated by the Amorphous algorithm. Ten nearest beacon nodes are considered to determine the coordinates of a single unknown node. The hop value and hop size of that unknown node are calculated from the 10 nearest beacon nodes, and by using the hop value and hop size, the weight is calculated. After the weight estimation, the weighted sum is calculated, and using the weighted sum, the coordinates of an unknown node are determined. Simulation results indicate that the proposed Weighted Centroid Amorphous method achieves superior accuracy, with rates of 90.41%, 89.57%, and 86.10% compared to the traditional Amorphous, improved Amorphous, and Ensemble approach, respectively.

Estimating Environmental Cost Throughout Model’s Adaptive Life Cycle

With the rapid increase in the research, development, and application of neural networks in the current era, there is a proportional increase in the energy needed to train and use models. Crucially, this is accompanied by the increase in carbon emissions into the environment. A sustainable and socially beneficial approach to reducing the carbon footprint and rising energy demands associated with the modern age of AI/deep learning is the adaptive and continuous reuse of models with regard to changes in the environment of model deployment or variations/changes in the input data. In this paper, we propose PreIndex, a predictive index to estimate the environmental and compute resources associated with model retraining to distributional shifts in data. PreIndex can be used to estimate environmental costs such as carbon emissions and energy usage when retraining from current data distribution to new data distribution. It also correlates with and can be used to estimate other resource indicators associated with deep learning, such as epochs, gradient norm, and magnitude of model parameter change. PreIndex requires only one forward pass of the data, following which it provides a single concise value to estimate resources associated with retraining to the new distribution shifted data. We show that PreIndex can be reliably used across various datasets, model architectures, different types, and intensities of distribution shifts. Thus, PreIndex enables users to make informed decisions for retraining to different distribution shifts and determine the most cost-effective and sustainable option, allowing for the reuse of a model with a much smaller footprint in the environment. The code for this work is available here: https://github.com/JEKimLab/AIES2024PreIndex

Exploring emergent microservice evolution in elastic deployment environments

Microservices have become an important technology to enable the dynamic composition of large-scale self-adaptive systems. Although modern microservice ecosystems provide a variety of autonomous adaptation mechanisms, when focusing on the microservice itself, they can only account for changes in the sheer increase in workload volume. On the other hand, when workload patterns change, efficient treatment requires the intervention of DevOps experts to manually evolve the internal architecture of services. Given the need to quickly adapt systems to respond to changes, solely relying on DevOps to react to workload pattern changes becomes a bottleneck for future systems. To address this issue, we advance the concept of emergent microservices, that autonomously adapt and evolve their internal architectural composition to better handle changes in the pattern of incoming requests without human intervention. We demonstrate the effectiveness of our approach by exploring this novel concept in the context of a microservice-based Smart City platform.Editor’s note: Open Science material was validated by the Journal of Systems and Software Open Science Board.

Design of voice command recognition chip based on heterogeneous acceleration

Abstract Speech, as one of the earliest forms of communication used by humans, can effectively convey information. However, the current deep neural network models for speech recognition are generally large in scale and can only be deployed in the cloud, which imposes high deployment environment requirements and power consumption, thereby limiting their implementation on embedded devices. In the context of end-to-end speech recognition, a series of challenges are encountered, including power consumption constraints, computing power limitations, network dependencies, privacy protection, bandwidth restrictions, and communication delays. To address these issues, this paper proposes the design of an end-to-end voice command recognition chip based on deep neural networks specifically for recognizing voice commands in specific scenarios. This chip achieves low power consumption and minimal delay in recognition. Additionally, we introduce a weighted, overloadable chip architecture to enable seamless scene migration, ultimately aiming to resolve the aforementioned challenges.

From code sharing to sharing of implementations: Advancing reproducible AI development for medical imaging through federated testing

BackgroundThe reproducibility crisis in AI research remains a significant concern. While code sharing has been acknowledged as a step toward addressing this issue, our focus extends beyond this paradigm. In this work, we explore “federated testing” as an avenue for advancing reproducible AI research and development especially in medical imaging. Unlike federated learning, where a model is developed and refined on data from different centers, federated testing involves models developed by one team being deployed and evaluated by others, addressing reproducibility across various implementations. MethodsOur study follows an exploratory design aimed at systematically evaluating the sources of discrepancies in shared model execution for medical imaging and outputs on the same input data, independent of generalizability analysis. We distributed the same model code to multiple independent centers, monitoring execution in different runtime environments while considering various real-world scenarios for pre- and post-processing steps. We analyzed deployment infrastructure by comparing the impact of different computational resources (GPU vs. CPU) on model performance. To assess federated testing in AI models for medical imaging, we performed a comparative evaluation across different centers, each with distinct pre- and post-processing steps and deployment environments, specifically targeting AI-driven positron emission tomography (PET) imaging segmentation. More specifically, we studied federated testing for an AI-based model for surrogate total metabolic tumor volume (sTMTV) segmentation in PET imaging: the AI algorithm, trained on maximum intensity projection (MIP) data, segments lymphoma regions and estimates sTMTV. ResultsOur study reveals that relying solely on open-source code sharing does not guarantee reproducible results due to variations in code execution, runtime environments, and incomplete input specifications. Deploying the segmentation model on local and virtual GPUs compared to using Docker containers showed no effect on reproducibility. However, significant sources of variability were found in data preparation and pre-/post- processing techniques for PET imaging. These findings underscore the limitations of code sharing alone in achieving consistent and accurate results in federated testing. ConclusionAchieving consistently precise results in federated testing requires more than just sharing models through open-source code. Comprehensive pipeline sharing, including pre- and post-processing steps, is essential. Cloud-based platforms that automate these processes can streamline AI model testing across diverse locations. Standardizing protocols and sharing complete pipelines can significantly enhance the robustness and reproducibility of AI models.

The Urban Deployment Model: A Toolset for the Simulation and Performance Characterization of Radiation Detector Deployments in Urban Environments.

Static and mobile radiation detectors can be deployed in urban environments for a range of nuclear security applications, including radiological source search-and-tracking scenarios. Modeling detector performance for such applications is challenging, as it does not depend solely on the detector capabilities themselves. Many factors must be taken into consideration, including specific source and background signatures, the topology and constraints of the deployment environment, the presence of nuisance sources, and whether detectors are mobile or static. When considering the simultaneous deployment of multiple, heterogeneous detectors, assessment of the system-wide performance requires the simulation of the individual detectors, and a system-level analysis of the detection performance. In radiological source search-and-tracking scenarios, performance is mostly dominated by the probability of encounter, which depends on the specifics of a given deployment, e.g., static vs. mobile detectors or a combination of both modalities, the number of detectors deployed, the dynamic vs. static setting of false alarm rates, and individual vs. networked operation. The Urban Deployment Model (UDM) toolset was specifically developed to cover the gap in the available generic frameworks for the simulation of radiation detector deployments at city scales. UDM provides a unified and modular framework to support the simulation and performance characterization of heterogeneous detector deployments in urban environments. This paper presents the key components along the UDM workflow.

End-to-end elevator voice command recognition system based on improved convolutional neural network

Abstract As a common life scenario, the use of voice control in elevator operation can effectively reduce public health risks and improve user experience in the post-epidemic era. However, due to the characteristics of its closure and public space, it is difficult to deploy the elevator. To solve this problem, an end-to-end speech command recognition algorithm based on a new deep neural network is proposed. The algorithm uses the self-built corpus for training, and uses the speech segment interception algorithm to obtain audio segments under the premise of speech stream as input, puts them into the network model in real time for reasoning, and drives the elevator to run according to the output results. The results show that compared with the recognition model using MFCC-CNN, the network has achieved about 10% improvement in accuracy under the premise of smaller computation, and compared with the DTW algorithm, the network has improved the accuracy by about 25%. Finally, a typical deployment environment is constructed to prove the correctness and effectiveness of the method in practical application.

[formula omitted] : An adaptive data aggregation algorithm for IoT infrastructure

In IoT infrastructure, high-frequency sensing and subsequent transmission of sensed data to computational facilities can lead to redundant data storage and processing, consuming significant storage and processing capacity. As a result, the IoT infrastructure needs more data transmission cycles, leading to data redundancy and low network up-time due to the drainage of limited battery capacity. Conversely, if the data is communicated at a lower rate, it may cause absolute data delivery to the processing unit, which is useless. As a result, a well-designed data aggregation algorithm is required. This paper proposes the ADA2−IoT, an Adaptive Data Aggregation Algorithm for IoT Infrastructure tailored to optimize parameters such as low data redundancy, limited data communication cycles, and high IoT infrastructure up times. The proposed algorithm consists of two key components: the Route Data Aggregator (RDA) performs aggregation during data transit towards the Edge node or gateway, and the Node Data Aggregator (NDA) performs data aggregation during capturing or sensing data. The algorithm employs metrics like Age of Information (AoI) and data freshness factor during the node and route data aggregation phase to capture and timely deliver data to the Edge node, where this data is processed for informed decision-making. The proposed algorithm was efficiently tested on a simulation and IoT hardware deployment environment. Both simulation and hardware results demonstrate a substantial improvement in QoS parameters, such as a decrease in data redundancy and packet exchanges, leading to considerable energy savings and prolonging the lifespan of IoT infrastructure.

EROI development and validation of a framework to assess the return on the environment of RFID deployment

This study is grounded in the growing significance of environmental sustainability and the widespread adoption of RFID technology across various industries and is aimed to explore the influence of RFID implementation in supply chains by developing a tool that calculates the net balance of CO2 annual emissions. The tool, known as “Return on the Environment” (EROI), is based on a widely accepted environmental assessment method and it calculates the Global Warming Potential (GWP100) incurred and avoided at various stages in the supply chain strictly related to RFID technology introduction. To validate the tool, two RFID deployments have been assessed: one involving a pharmaceutical product tagged on its secondary packaging to monitor the flow of products through the distribution channel, and another a medical device tagged on both primary and secondary packaging to oversee product flow from the supplier distribution center to the hospital operating theatres. In both cases, the results indicate that implementing RFID technology reduced GWP100 compared to the scenarios without RFID. This was primarily due to decreased product shrinkage, lower missing or expired products, and reduced additional transportation due to shipping errors. The tool is versatile and it could be applied to any type of product, serving as a source of inspiration for those who want to assess the sustainability of RFID technology not only from an economic perspective, i.e. ROI calculation, but also from an environmental view. Future work will address the third level of sustainability, RFID social sustainability, that is the impact RFID deployments may have on empowering people, improving staff and employees working conditions, and creating possibilities for high-value job opportunities.

Prospects and Challenges of Renewable Energy in Ethiopia With Special Focus on Trends and Practices

This study aimed to explore the prospects and challenges of Renewable Energy in Ethiopia with a special focus on trends and practices. The study was conducted using a qualitative approach to understand the underlying problems of the topic holistically. The targets of the study included the Ministry of Water and Energy, the Ethiopian Energy Authority, the Ethiopia Water Technology Institute, and Ethiopian Solar Development Associations, which are responsible for leading the Renewable Energy sector of the country. A purposive sampling technique was used to collect data from sample respondents with rich experiences and exposure to Ethiopia's renewable energy sector's practices and trends through an in-depth interview, Focus Group Discussions (FGD), and document analysis of Renewable energy policy and strategy and Energy policy documents. The data were analyzed using thematic approaches. The findings revealed that Renewable Energy in Ethiopia has suffered from a lack of infrastructure and investment in renewable energy projects, technical and financial constraints in developing renewable energy technologies, regulatory and policy barriers, limited access to financing, and Intermittent nature of some renewable energy sources and challenges for grid integration and stability. Lack of strong and committed leadership is also a core problem in supporting the sector. Based on these findings, the researcher suggested that the leadership on renewable energy in Ethiopia needs to be characterized by a commitment to policy and regulatory framework implementations that support the sector's sustainable development. By formulating clear and supportive policies, the government seeks to create an enabling environment for renewable energy deployment, attract investment, and ensure the long-term viability of renewable energy projects. Overall, in alignment with the government's commitment to advance renewable energy concerning the country's broader development objectives and underscoring the transformative potential of clean energy in driving Ethiopia's sustainable growth and prosperity; the country’s leadership perspective on renewable energy needs to reflect a comprehensive and multifaceted approach that encompasses energy access, environmental sustainability, economic development, technological innovation, regional cooperation, affordability, and policy support.

Combined multi-level context aggregation and attention mechanism method for photovoltaic panel extraction from high resolution remote sensing images

ABSTRACT In the context of global carbon emission reduction, solar photovoltaic (PV) technology is experiencing rapid development. Using high-resolution remote sensing images to accurately obtain PV information over a large region, including location and size, has the advantages of high statistical efficiency and timely data update for the PV energy management. Due to the intra-class diversity of PV panels and the intricate variability in their deployment environments, existing semantic segmentation methods often have problems such as under-segmentation and mis-segmentation. To alleviate these problems, this paper proposes an improved DeepLabv3+ semantic segmentation network to more accurately extract PV panels from high-resolution remote sensing images. With the aim of alleviating under-segmentation, a multi-level context aggregation module is developed. This module can enhance the model’s ability to learn the characteristics of PV panels and their surrounding environment by aggregating rich contextual information from multi-scale and semantic levels. To alleviate the problem of mis-segmentation, a hybrid attention module is introduced. This module sequentially and adaptively adjusts the weight distribution in both the channel and spatial dimensions, thus enabling the model to focus more on the feature information and spatial positions of PV objects. Experiments conducted on a self-constructed Beijing PV segmentation dataset show that the method in this paper has advantages of completeness and accuracy in extracting PV panels compared to the baseline model and current mainstream semantic segmentation network. In addition, the results of experiments on extracting PV panels in real region show that our model also has good stability and generalization capability.

A Lightweight Secure Scheme for Underwater Wireless Acoustic Network

Due to the open underwater channels and untransparent network deployment environments, underwater acoustic networks (UANs) are more vulnerable to hostile environments. Security research is also being conducted in cryptography, including authentication based on asymmetric algorithms and key distribution based on symmetric algorithms. In recent years, the advancement of quantum computing has made anti-quantum attacks an important issue in the field of security. Algorithms such as lattice and SPHINCS+ have become a research topic of interest in the field of security. However, within the past five years, few papers have discussed security algorithms for UANs to resist quantum attacks, especially through classical algorithms. Some existing classical asymmetric and symmetric algorithms are considered to have no prospects. From the perspective of easy deployment in engineering and anti-quantum attacks, our research focuses on a comprehensive lightweight security framework for data protection, authentication, and malicious node detection through the Elliptic Curve and Hash algorithms. Our mechanism is suitable for ad hoc scenarios with limited underwater resources. Meanwhile, we have designed a multi-party bit commitment to build a security framework for the system. A management scheme is designed by combining self-certifying with the threshold sharing algorithm. All schemes are designed based on certificate-less and ad hoc features. The proposed scheme ensures that the confidentiality, integrity, and authentication of the system are well considered. Moreover, the scheme is proven to be of unconditional security and immune to channel eavesdropping. The resource and delay issues are also taken into consideration. The simulations considered multiple variables like number of nodes, attackers, and message length to calculate proper values that can increase the efficiency of this scheme. The results in terms of delay, delivery ratio, and consumption demonstrate the suitability of the proposal in terms of security, especially for malicious node detection. Meanwhile, the computational cost has also been controlled at the millisecond level.

M3BAT: Unsupervised Domain Adaptation for Multimodal Mobile Sensing with Multi-Branch Adversarial Training

Over the years, multimodal mobile sensing has been used extensively for inferences regarding health and well-being, behavior, and context. However, a significant challenge hindering the widespread deployment of such models in real-world scenarios is the issue of distribution shift. This is the phenomenon where the distribution of data in the training set differs from the distribution of data in the real world---the deployment environment. While extensively explored in computer vision and natural language processing, and while prior research in mobile sensing briefly addresses this concern, current work primarily focuses on models dealing with a single modality of data, such as audio or accelerometer readings, and consequently, there is little research on unsupervised domain adaptation when dealing with multimodal sensor data. To address this gap, we did extensive experiments with domain adversarial neural networks (DANN) showing that they can effectively handle distribution shifts in multimodal sensor data. Moreover, we proposed a novel improvement over DANN, called M3BAT, unsupervised domain adaptation for multimodal mobile sensing with multi-branch adversarial training, to account for the multimodality of sensor data during domain adaptation with multiple branches. Through extensive experiments conducted on two multimodal mobile sensing datasets, three inference tasks, and 14 source-target domain pairs, including both regression and classification, we demonstrate that our approach performs effectively on unseen domains. Compared to directly deploying a model trained in the source domain to the target domain, the model shows performance increases up to 12% AUC (area under the receiver operating characteristics curves) on classification tasks, and up to 0.13 MAE (mean absolute error) on regression tasks.