Vehicle trajectory prediction in Internet-of-Vehicles requires collaborative learning over sensitive trajectories under intermittent connectivity and partially trusted participants. ChainDrive-FL-VRA coordinates semi-asynchronous federated learning on a permissioned consortium ledger using Practical Byzantine Fault Tolerance (PBFT), while keeping raw trajectories and raw model-update tensors off-chain. Each client submits an on-chain header containing a commitment and hash of the local update, together with zero-knowledge proofs that certify [Formula: see text]clipping and anchor-consistency. Validators admit only proof-checked updates, compute staleness- and reputation-aware robust weights, and publish a proof of correct aggregation that binds the aggregation commitment and the committed global model hash to the admitted committed updates under fixed-point weights. A contextual-bandit trigger selects aggregation timing under client churn. Experiments on NGSIM US-101 and I-80 show improved ADE/FDE/RMSE and improved robustness under staleness and anomalous updates, while on-chain artifacts remain at kilobyte scale per update and per aggregation event.
The fast communication networks are crucial to support the modern digital services like cloud computing, massive data transmissions and real time multimedia applications. Since network traffic is constantly increasing exponentially, a proper approach to managing congestion is required to ensure the delivery of information is stable, minimize delays, and efficiently use bandwidth. Conventional congestion control mechanisms tend to use systems that are based on fixed rules and thresholds, and may be unable to be flexible in highly dynamic network situations. A graphical congestion control model is intelligent based on a multi-agent Deep Q-Learning model in which the distributed agents are tasked with monitoring network conditions such as queue length, delay, packet loss, and available bandwidth. The agents are taught the best acting policies in traffic regulation by means of interaction with their network environment and dynamically change their rates of transmission to reduce congestion. Learning organization is decentralized and enhances adaptability and scalability within large network systems. In comparison to traditional methods that attained a throughput of 780-910 Mbps, 2.9-5.8% packet loss, and 84-120 ms end-toend delay, performance evaluation has shown to achieve better network performance of 960 Mbps throughput, 1.8 end-to-end delay, and 0.8% end-to end packet loss. These enhancements underscore the success of smart use of reinforcement learning methods in adaptive congestion control in high-speed networking settings.
With the acceleration of urbanization and the promotion of the “dual carbon” goal, the road transport system is facing the triple challenges of efficiency bottlenecks, excessive carbon emissions, and data security risks. In view of the shortcomings of the existing research in dynamic response, multi-objective collaboration and privacy protection, this paper proposes a three-in-one intelligent management framework: (1) construct a real-time dynamic path optimization model based on Deep Reinforcement Learning (DRL), and realize the precise regulation of traffic flow through multi-source data fusion and adaptive reward mechanism; (2) Design a multi-objective optimization model integrating carbon trading mechanism to quantify the synergistic relationship between transportation efficiency, carbon emissions and economic costs; (3) Develop a distributed data management framework based on blockchain, and use zero-knowledge proof and smart contract technology to protect user privacy. The peak simulation experiment based on the fifth ring road section of Beijing shows that the proposed method reduces the average traffic time by 18.7%, the carbon emission by 23.5%, and the risk of data leakage by 76% compared with the traditional algorithm. This study provides theoretical and technical support for the construction of a safe, efficient and low-carbon intelligent transportation system.
This paper proposes Temporal Convolutional Networks (TCNs) for cryptocurrency forecasting at minute-resolution. TCNs show better accuracy to XGBoost, LightGBM, and LSTM. However, TCNs are less robust than the tree models regarding volatility. While TCNs require much more computations at inference than LightGBM, they run faster than LSTMs. The performance of TCNs is best configured using a TCN with dilated convolutions to capture the temporal patterns, and with residual connections for the stability.
As the climate crisis intensifies and urban populations swell, megacities face compounding threats from carbon emissions, urban heat islands (UHIs), and ecosystem degradation. While nature-based solutions (NbS) offer a promising response through ecological restoration and carbon sequestration, current NbS deployments are often fragmented, non-adaptive, and lack quantitative optimization. This research presents a cutting-edge, artificial intelligence (AI)-driven architecture that operationalizes NbS through a scalable, data-intensive framework. It integrates deep learning (DL) for satellite-derived land classification, graph neural networks (GNNs) for spatial co-benefit mapping, and reinforcement learning (RL) with dynamic reward weighting to optimize intervention strategies in real time. Life cycle assessment (LCA) and ecosystem service valuation modules are embedded to ensure holistic, cross-sectoral impacts. The architecture is deployed in a high-resolution case study of Dhaka, Bangladesh, a climate-vulnerable megacity, achieving over 8,500 metric tons of modeled annual carbon sequestration, 2.1°C reduction in UHI intensity, and quantifiable gains in urban biodiversity and flood mitigation. The system ingests multi-source data, including Sentinel-2, LiDAR, and CMIP6 climate projections, while leveraging federated learning to ensure decentralized, privacy-preserving optimization across municipal zones. A carbon market compatibility layer, aligned with Verra, UN-REDD+, and Article 6 frameworks, enables eligibility for climate finance and offsets. The approach also integrates social equity metrics and indigenous ecological knowledge to prioritize interventions in marginalized zones. This work delivers a first-of-its-kind decision-support platform for AI-optimized NbS that is globally scalable, policy-aligned, and climate-finance ready. It represents a paradigm shift from heuristic-based planning to algorithmically adaptive ecosystem engineering, accelerating progress toward net-zero emissions, SDG convergence, and resilient urban futures. The framework is poised to inform urban sustainability strategies worldwide, offering a replicable model for AI-governed environmental transformation in the age of planetary emergency.
The demand for privacy-preserving machine learning has led to the rise of Federated Learning (FL), where multiple clients collaboratively train a model without sharing raw data. Despite its privacy benefits, FL is vulnerable to Byzantine failures, where malicious or faulty participants inject corrupted updates, threatening model integrity. To address this, a range of Byzantine-resilient aggregation techniques have been proposed, including statistical filters (e.g., Trimmed Mean, Krum), trust-based weighting, cryptographic protocols, and hybrid strategies. This paper presents a systematic literature review (SLR) of these defenses, evaluating their robustness, scalability, and suitability for real-world applications. Challenges such as non-IID data, adaptive attacks, and trade-offs between security and efficiency are critically examined. In addition, we explore emerging trends such as domain-specific defenses, energy-aware FL, quantum-resilient methods, and federated zero-knowledge proofs. A novel classification of hybrid approaches and a standardized benchmarking framework are proposed to guide future research. This review aims to support the development of resilient, efficient and scalable decentralized learning systems in adversarial environments.
Bitcoin has attained increasing recognition and interest from individuals and corporations, with more than $1 billion market capitalization. Twitter users’ sentiment on the topic is a major factor that influences volatility of Bitcoin’s price. Compared to other financial markets, there are a limited number of studies that discuss the price fluctuation prediction of Bitcoin using Twitter sentiment. A dataset with 16 million tweets from August 2018 to October 2019 was utilized for finding the correlation between the daily close price of Bitcoin and Twitter sentiment. This dataset was pre-processed by following steps such as removing null, duplicate and non-English tweets. The sentiment analysis was carried out using VADER sentiment analyzer. This research utilized hyperparameter optimization and improved two deep learning models (with Long Short-Term Memory and Convolutional Neural Network architectures), for the tasks of direction and magnitude prediction with accuracies of 82.35% and 72.06%, respectively on test datasets. With hyperparameter optimization this research addresses a gap in the existing research of this research area, which was not utilizing hyperparameter optimization to improve deep learning models.
Montaser N.A. Ramadan, Mohammed A. H. Ali, Hadi Jaber, Mohammad Alkhedher
Air pollution in industrial zones significantly impacts environmental safety and worker health. This paper presents a novel decentralized IoT-federated learning (FL) framework, uniquely integrated with blockchain security, designed to provide a mechanistic understanding and accurate predictive modeling of air pollutant exposure in industrial environments. The novelty lies in the integration of a hybrid EMD-Transformer-BiLSTM prediction model with a blockchain-backed federated learning mechanism, providing secure, tamper-proof decentralized model updates. Three IoT-based sensing units, deployed across an industrial facility for five months, continuously monitored pollutants (PM2.5, PM10, CO₂, VOCs, CH₂O, CO, and O₃) and environmental factors (temperature, humidity). The innovative model improved prediction accuracy from 83.12 % to 92.5 % for short-term (5-minute) forecasts, stabilizing at 84.7 % for 60-minute predictions after 15 FL rounds. Model validation indicated strong predictive reliability (R² = 0.89), significantly reducing prediction errors (Mean Absolute Error and Root Mean Square Error). Blockchain integration successfully ensured data integrity, identifying and rejecting over 98.7 % of unauthorized updates. Additionally, a swarm intelligence approach optimized decentralized model aggregation, minimizing communication overhead despite increased security latency (FL rounds increased from 7.5 s to 13.5 s for 500 clients). Real-time RGB-based air quality index visualization and cloud-based spatio-temporal mapping provided actionable insights into pollutant dynamics. This study demonstrates a distinct advancement in air pollution monitoring by combining federated learning, blockchain technology, and real-time adaptive visualization for enhanced environmental safety in industrial settings.
May 7, 2025·2025 2nd International Conference on Research Methodologies in Knowledge Management, Artificial Intelligence and Telecommunication Engineering (RMKMATE)
S. Neelavathy Pari, M. D. Anto Praveena, Ms. S. Kaaviya, Ms. S. Kanishka
The Internet of Things (IoT) has transformed various industries by enabling seamless connectivity among smart devices, but its open nature exposes it to security vulnerabilities. Blockchain technology, with its decentralized and immutable properties, offers a promising solution to enhance IoT security. However, existing anomaly detection approaches in IoT networks face limitations such as high false positives, scalability issues, and lack of real-time threat mitigation. To address these challenges, this research integrates the Isolation Forest algorithm with a blockchain-based smart contract for efficient anomaly detection. The Isolation Forest algorithm is used to classify network anomalies by classifying normal and abnormal traffic patterns, while smart contracts detect anomalies in real-time network traffic, ensure data integrity, automate threat responses, and provide tamper-proof logging. Experimental evaluations demonstrate the effectiveness of this approach, achieving improved accuracy of 95 percent along with other measures such as precision, recall, and F1-score also achieving good results compared to other traditional methods. The proposed framework enhances IoT security by reducing false alarms, increasing detection sensitivity, and enabling real-time threat identification, making it a scalable and robust solution for modern IoT environments.
Recently, with the gradual development of machine learning technology, more and more people are trying to apply machine learning technology in various fields, and finance is one of the important fields. This work investigates the optimization of cryptocurrency portfolios by combining Long Short-Term Memory (LSTM) time series forecasting with traditional portfolio optimization methods. The focus of the paper is on using the historical price data from the past six years of Bitcoin (BTC), Ethereum (ETH), and Litecoin (LTC) to train LSTM models, which are then used to predict the prices of these cryptocurrencies for the period from January to June 2024. These predictions are subsequently incorporated into an extended Markowitz framework to optimize the portfolio on a monthly basis. The results indicate that the LSTM-enhanced portfolio optimization method yields higher returns and better risk management compared to traditional methods. This finding could prove that it is feasible and effective to apply machine learning methods, especially time series forecasting methods, to cryptocurrency portfolios.
As urban populations continue to grow and technological advancements reshape our world, the concept of smart cities has emerged as a beacon of innovation in urban development. This chapter explores the transformative potential of two cutting-edge technologies, deep learning and blockchain, in the context of smart cities. Deep learning, a subset of artificial intelligence, offers powerful tools for analyzing vast amounts of urban data to optimize various aspects of city life, including traffic management, public safety, energy usage, waste management, and healthcare. Concurrently, blockchain technology provides decentralized and secure platforms for transactions, supply chain management, identity verification, smart contracts, and data privacy, addressing critical challenges faced by modern cities. By examining the applications of deep learning and blockchain in smart cities, this chapter delves into the synergistic possibilities and challenges of integrating these technologies into urban infrastructure. Real-world case studies highlight successful implementations and lessons learned, while future perspectives envision the continued evolution of smart cities powered by deep learning and blockchain innovations. As a result, this chapter emphasizes the critical importance of collaboration, innovation, and strategic planning in unlocking the vast potential of deep learning and blockchain technologies to revolutionize smart cities and improve urban life on a global scale. By fostering partnerships, driving innovation, and implementing thoughtful strategies, cities can harness the transformative power of these technologies to create more efficient, sustainable, and resilient urban environments for the benefit of all residents.
G. Revathy, R. Indhumathi, A. Nisha Jebaseeli, Abdul Azis Fairosebanu · 6 authors
Distributed ledger technology, or blockchain, allows transactions to be recorded on a network of computers in a transparent and safe manner. Every transaction, or “block,” is cryptographically connected to the one before it, creating a chain of blocks that, once recorded, is unchangeable or manipulable. Blockchain’s irreversible and decentralized structure guarantees transaction transparency and confidence without the need for middlemen. The creation of smart cities can greatly benefit from blockchain technology, but there are a number of issues that need to be taken into account, including scalability, interoperability, regulatory compliance, energy usage, and privacy concerns. In order to create standards, protocols, and governance frameworks that are specifically suited to the requirements of smart cities, government organizations, business stakeholders, and technology providers must work together for successful implementation. One area of artificial intelligence that has great potential to improve many facets of smart city development is deep learning. Deep learning can help create more responsive, sustainable, and efficient urban systems by using deep neural networks to evaluate massive volumes of data, identify trends, and make predictions. Blockchain and deep learning have a lot of potential for use in smart city applications, but there are a few issues that need to be resolved first. These issues include data security and privacy, algorithm bias and fairness, interpretability and explainability, scalability and resource limitations, and legal compliance. To solve these issues and guarantee the ethical and responsible application of deep learning in smart city development, cooperation between government organizations, business players, academic institutions, and technology companies is also necessary for successful implementation.
Cloud computing has become a critical component of modern IT infrastructure, offering businesses scalability, flexibility, and cost efficiency. Unoptimized cloud migration strategies can lead to significant financial waste due to inefficient resource allocation, redundant workloads, and unpredictable cloud expenses. Traditional methods often rely on static provisioning and manual decision-making, leading to suboptimal cloud resource utilization. This research introduces an AI-driven framework for intelligent cloud planning and migration aimed at reducing cloud costs while maintaining high performance and compliance standards. The proposed framework leverages machine learning (ML), deep learning (DL), and reinforcement learning (RL) techniques to automate workload distribution, real-time scaling, and dynamic cost optimization. It integrates Predictive Analytics Engine: Uses AI models (Long Short-Term Memory LSTMs, CNNs, and Transformers) to analyze historical workload data and forecast future resource demands. Optimization Algorithm: Implements AI-driven cost minimization functions, optimizing resource allocation while maintaining Quality of Service (QoS). Automated Migration Engine: Reduces manual intervention by executing AI-based cloud workload transfers efficiently. Security and Compliance Module: Uses explainable AI (XAI) and federated learning to maintain cloud security, privacy, and regulatory compliance. A proof of concept (PoC) is developed and evaluated across multiple cloud platforms (AWS, Azure, Google Cloud) with real-world datasets. Experimental results indicate that the AI-driven framework achieves: Cost savings of up to 42% compared to traditional cloud migration strategies. Resource utilization improvement by 53%, ensuring minimal wastage. Reduction in system downtime by 75%, leading to higher reliability. Reduction in manual intervention by 85%, automating resource scaling and load balancing. The research paper also presents real-world case studies across finance, healthcare, e-commerce, and manufacturing sectors, demonstrating the tangible impact of AI-based cloud optimization. This research explores future advancements in cloud computing, including Quantum AI for cloud workload acceleration, Blockchain for transparent cloud cost auditing, and Decentralized AI governance for multi-cloud management. This study contributes to the growing field of AI-driven cloud cost optimization, providing a roadmap for enterprises, cloud architects, and AI researchers to achieve cost-efficient, high-performance, and automated cloud management.
In recent years, various digital currencies have emerged, among which Bitcoin has been widely accepted as an alternative to sovereign currencies for commodity trading. However, the dramatic volatility of bitcoin prices can pose a risk to global financial markets. In this paper, we firstly construct a more comprehensive forecasting index system from seven aspects, and then construct a VMD-GRU model. This model uses the variational modal decomposition (VMD) to decompose the time series into intrinsic mode functions (IMFs) and use gated recurrent unit (GRU) to forecast different IMFs. This paper also compares the forecast results with classical machine learning models and deep learning models, and the results show that the forecast accuracy of the VMD-GRU model is more than 16% better than other models.
Bin Wang, Zhao Tian, Fengxiao Tang, Heng Pan · 6 authors
Vehicle trajectory prediction plays a crucial role in IoT-based intelligent transportation systems, which can effectively address key issues, such as driving safety and multivehicle collaboration. However, the sensitivity of trajectory data and the reluctance of data holders to share it constrain the prediction model’s ability to capture vehicle behavior patterns in different scenarios. To address the above problems, we propose a blockchain-enabled asynchronous federated proximal policy optimization framework (BE-AFPPO) for the trajectory prediction of self-driving vehicles. First, we propose a curiosity proximal policy optimization (C-PPO) algorithm. The method utilizes a driven exploration strategy to actively motivate the intelligent agent to explore the unknown state space. The avoidance policy model reaches a local optimum when processing trajectory data. In addition, we design historical gated recurrent unit (GRU) and future GRU as input layers. The target’s historical motion features and future trajectory features are extracted, respectively. Then, various data is received through asynchronous federated learning. This model can fully learn the vehicle’s behavior patterns in different scenarios, which improves prediction accuracy. Based on this, we develop a blockchain-based dynamic group practical Byzantine fault tolerance (DG-PBFT) consensus algorithm. This enhances the credibility and integrity of the data while enriching the sources of trajectory data. Finally, we perform the experiments on the publicly available dataset nuScenes. The results demonstrate that the proposed method improves the robustness and accuracy of trajectory prediction.
Stanislav I. Trofimov, Leonid Voskov, Mikhail Komarov
The development of intelligent transportation systems (ITSs) is penetrating many economies around the globe. This paper presents three key innovations in the field of intelligent transportation systems, as follows: (1) a novel tokenization approach where each vehicle is represented as a macro-token subdivided into 500,000 micro-tokens for precise condition monitoring, (2) a comprehensive mathematical model for vehicle state assessment incorporating multiple operational factors, and (3) the GDEPZ method for optimizing data transmission via satellite communication. These innovations enable the autonomous control of technical conditions, transparent fleet management, and efficient data processing in hard-to-reach areas. Various researchers in both industry and academia are looking into more efficient management methods for both vehicles and related data processing aspects. A vast trend related to the latter is the distributed data processing of transmitted data. This article discusses approaches to the use of blockchain technology in ITSs. It explores the use of blockchains in modern transport industries. In particular, the paper proposes a novel approach to the maintenance of public transportation vehicles and buses. The specificity of the proposed approach is the autonomous control of technical conditions using information systems. When using blockchain technology, building a transparent vehicle fleet management system is possible. The specificity of the proposed approach lies in data processing. Within the organization, confidence in data increases, the possibility of manipulating transportation is eliminated, and the decision-making chain is reduced. As a result, the system can manage itself. This also helps to increase the service life of vehicles, makes it possible to predict their malfunctions, and improves the quality of data on their technical conditions.
The modern concept of taxation has been accelerated by the emergence of internet-based economy and the use of cryptocurrencies. This shift raises various difficulties for tax authorities in terms of revenue estimates; it requires sophisticated methods for quantitative analysis of intricate economic trends. Previous works have mainly employed conventional ML methods which are weak in their ability to recognize dependencies in the features for data with high dimensionality, hence poor forecast precision. To counter these drawbacks, the use of CapsNets with CNNs, which forms a combined model contributing to better predictive capability is introduced. CapsNets ability to keep ownership of spatial hierarchies and sophisticated features of the input makes the method to accurately extract and analyze features from digital transaction data. Working in Python, the suggested model was compared to several traditional algorithms with which it demonstrated the highest accuracy rate at 99.1%. As concluded from the evaluation, the CapsNet-CNN model is not only resistant to the formulation of new structures, but also flexible to the fluidity of the digital economy environment. This work demonstates the promise of using modern DL methodologies for improving the accuracy of the tax revenues and provides insights for the policy makers who are keen on evoking sensitive and dynamic response strategies of the taxation authorities in a world where more and more of the economic operations are being carried out in the cyberspace. Future research will quantify the components in the model and expand the investigation to other industries and locations.
A smart city is a fast-moving terrain that requires efficient and smart security mechanisms with resilience for solving the intricate challenges of modern urbanism. The current paper presents the critical review of machine learning, blockchain, and deep learning models in strengthening security and making the urban environment at smart cities more resilient. Not all the review articles heretofore successfully integrated these three pivotal domains, which lowered their practical applicability and insight depth. This paper reviews recent state-of-the-art models, including supervised and unsupervised machine learning algorithms, blockchain frameworks, and deep advanced learning architectures. The important machine learning models reviewed include Random Forest, Support Vector Machines, and K-means clustering. These were chosen for their already proved effectiveness in anomaly detection, predictive analytics, and classification tasks. On the other hand, blockchain models of Ethereum and Hyperledger Fabric would be evaluated for decentralized security features, immutability, and capability to improve data integrity and transparency. Deep learning models can handle large-scaled, unstructured data and extract complex patterns. Their integration, therefore, shows great promise in enhancing the security and resilience of smart cities. This paper contributes to a holistic, multidisciplinary viewpoint, filling literature gaps by providing a basic framework for further research and implementation of smart city initiatives. Such insights synthesized across these domains set a course for innovation in the formulation of solutions that fortify urban infrastructures against emerging threats and enhance overall urban resilience. (Abstract)
Antonio Jesús Chaves, Cristian Martín, Kwang Soon Kim, Adnan Shahid · 5 authors
Machine learning data privacy has been improved with Federated Learning approaches. However, some obstacles to guaranteeing traceability, openness, and participant contribution incentives prevent its widespread use. In this study, Ethereum blockchain technology is integrated into the data stream Kafka-ML framework, presenting a novel asynchronous and blockchain-based Federated Learning approach. By utilising Ethereum for transparent and auditable participant tracking, this integration overcomes some shortcomings such as auditability and model sharing reliability. Furthermore, Ethereum smart contracts allow for automatic reward distribution systems, which promote equitable incentive systems and increased involvement in the Federated Learning process. To demonstrate its potential, an extensive evaluation has been carried out on a wireless net-work technology detection use case. By improving transparency, traceability, and incentive structures of Federated Learning, it is expected to strengthen the robustness of flexible machine learning collaboration with data streams.
Bitcoin, as the most widely recognized cryptocurrency, has attracted significant global attention from businesses, consumers, and investors. This study introduces a hybrid model that integrates an online self-learning stochastic configuration network (OSL-SCN) with conformal prediction. The model autonomously adjusts its parameters in response to real-time data.Predictions from the OSL-SCN are refined through conformal prediction, which generates confidence intervals to enhance reliability. The results, using historical Bitcoin prices from Wikipedia, demonstrate that the combined approach enhances both prediction accuracy and reliability.
Blockchain-based technology has completely revolutionized the development of the Internet of Vehicles (IoV) framework. This has led to increasing blockchain-based Internet of Vehicles application over the last decade. However, challenges persist, including scalability, interoperability, and security issues. This paper first presents the state-of-the-art overview on IoV systems along with their applications. Then, we explore novel technologies, including blockchain-based IoV and machine learning-based IoV and highlight how the blockchain technology could be integrated with machine learning for intelligent transportation systems in the IoV ecosystem. This paper has shown the potential of machine learning integration in addressing the technical challenges in individual blockchain-based Internet of Vehicles applications.