Cross-silo Federated Learning (FL) enables multiple institutions to collaboratively train machine learning models without directly sharing sensitive data. However, despite the absence of explicit data sharing, significant security risks remain, such as gradient inversion attacks that reconstruct private data from local client updates. To mitigate these risks, Homomorphic Encryption (HE) has been widely explored, allowing computations on encrypted data without decryption. However, HE introduces substantial computational and communication overhead, limiting its practical feasibility. To address this challenge, this work presents SlidHE, an efficient technique that integrates packing and packet sparsification methods with various selection strategies to significantly reduce overhead while preserving privacy and maintaining model accuracy, even in heterogeneous (non-IID) data settings. Experimental results demonstrate that SlidHE outperforms existing methods by effectively balancing efficiency, security, and performance. The full implementation of SlidHE is available on GitHub1.1https://github.com/AllanMSouza/SlidHE

The rise of mobile devices and growing concerns about model privacy have posed significant challenges in distributed artificial intelligence, especially due to the heterogeneity of devices, leading to model generalization and resource management issues. Federated Learning (FL), a method where machine learning models are trained collaboratively by sharing only local parameters with an aggregation server, faces challenges in model convergence, optimization, and communication overhead due to this heterogeneity. This paper introduces FedSCCS, an FL-based framework designed for such heterogeneous settings. FedSCCS clusters devices based on the similarity of their models, allowing for efficient model aggregation and improved resource utilization. Our evaluation, set against established benchmarks, shows that FedSCCS achieves superior accuracy compared to existing methods, indicating a promising direction for scalable and tailored FL solutions.

Federated Learning (FL) has emerged as a privacy-preserving powerful tool in decentralized Machine Learning (ML) environments. However, real-world scenarios often face bandwidth limitations that can be overwhelmed when all clients simultaneously perform training and communicate with the server in a federated system. Consequently, selection mechanisms are critical for identifying optimal subsets of clients to participate in the federation. Traditional selection methods, however, typically do not allow clients the autonomy to decide whether or not to contribute to the federation. Therefore, this paper proposes LetsFed, a client selection framework that respects client independence throughout the training process. The LetsFed framework differentiates between participating and non-participating clients, employing targeted selection mechanisms to address system challenges effectively. Empirical results demonstrate that LestFed can outperform, in dynamic client participation environments, literature solutions by up to 40%, reducing unnecessary data transmission by as much as 29%, while also enhancing the efficacy of the selection process.

Federated Learning (FL) allows distributed devices, known as clients, to train Machine Learning (ML) models collaboratively without sharing sensitive data. A characteristic of FL for mobile and IoT environments is system heterogeneity among clients, which can vary from low-end devices with constrained communication and computing resources to powerful devices with high-speed network access and dedicated GPUs. As the server must wait for all the clients to communicate their updates, slow clients (a.k.a. stragglers) will significantly increase the training time. To tackle this problem, we propose FedPulse, a Partial Training (PT) based mechanism to mitigate the effect of stragglers in FL. The idea is to reduce the training time by dynamically allocating smaller submodels to resource-constrained clients. Experimental results on famous classification datasets show that the proposed solution outperforms other submodel allocation mechanisms and reduces the training time by up to 58% with an accuracy loss of less than 1% when compared to FedAvg.

Federated Learning is a distributed approach in which multiple devices collaborate to train a shared global model. During its training, client devices must communicate their gradients to update the global model. This incurs significant communication costs (bandwidth utilization and number of messages exchanged), leading to many challenges (communication bottlenecks and scalability issues). Furthermore, the heterogeneous nature of clients’ datasets poses an extra training challenge. In this sense, we introduce FedCCSKD, a Federated Clustered Client Selection and Knowledge Distillation training algorithm, to decrease the overall communication costs. FedCCSKD is an innovative combination of: (i) client selection, and (ii) knowledge distillation approaches with three main objectives: (i) reducing the number of devices training at every round; (ii) increasing convergence speed; and (iii) mitigating the effect of clients’ heterogeneous data on the global model effectiveness. Our experimental evaluations on MNIST and MotionSense datasets demonstrate that FedCCSKD is highly efficient in training the global model until convergence. FedCCSKD reaches a higher accuracy score and faster convergence than state-of-the-art baseline models. Our results also show higher performance when analyzing the accuracy scores on the clients’ datasets.

The growing number of vehicles has led to increased emissions of polluting gases, necessitating accurate forecasting for effective mitigation strategies and sustainable urban development. Leveraging computational resources in vehicles, this study presents a framework, called EcoPredict, for predicting CO2emissions in collaborative vehicular network environments. The framework implements three forms of learning methods—centralized, federated, and split—using urban sensor networks for data collection. Experiments carried out in realistic vehicular mobility scenarios demonstrate the framework’s robustness and efficiency in providing real-time emission predictions. Each learning architecture has its own advantages and limitations regarding performance, training time, latency, communication overhead, and data privacy. Therefore, this work aims to assess their performance to analyze their effectiveness in urban environments.

Federated learning (FL) is a distributed approach that enables collaborative training of a shared machine learning (ML) model for a given task. FL requires bandwidth-demanding communication between devices and a central server, which is a cause of many issues such as communication bottlenecks and scaling in the network. Therefore, we introduce the CCS (Compressed Client Selection) algorithm aimed at decreasing the overall communication costs for fitting a model in the FL environment. CCS employs a biased client selection strategy that reduces the number of devices training the ML model and the number of rounds required to reach convergence. In addition, the compression method Count Sketch is implemented to reduce the overhead in client-to-server communication. A use case on the Human Activity Recognition dataset is performed to evaluate CCS and compare it with other state-of-the-art approaches. Experimental evaluations show that CCS efficiently reduces the overall communication overhead for fitting a model and its convergence in a FL environment. In particular, CCS reduces up to 90% the communication overhead compared to literature approaches while providing good convergence even in scenarios where the data are not-independently and identically distributed among client devices.

Vehicular Edge Computing (VEC) has emerged to offer cloud computing services closer to vehicular users by combining vehicles and edge computing nodes into Vehicular Clouds (VCs). In this scenario, an intelligent task scheduler must decide which VC will run which tasks, considering contextual aspects like vehicular mobility and tasks’ requirements. This is important to minimize both processing time and monetary costs. However, such direct optimization can lead to unfairness in resource usage, easily leading to (as we will show) decreased performance. Towards this end, in this work, we propose FARID, a task scheduling mechanism that considers contextual aspects of its decision process and applies a probabilistic selection function on VCs to balance the processing load and increase the fairness in the use of vehicular resources. Compared to state-of-the-art solutions, FARID has a higher level of fairness and can schedule more tasks while minimizing monetary costs and system latency.

Vehicular Edge Computing (VEC) is a promising paradigm that provides cloud computing services closer to vehicular users. In VEC, vehicles and communication infrastructures can form pools with computational resources to meet vehicular services with low-latency constraints. These resource pools are known as Vehicular Cloud (VC). The usage of VC resources requires a task scheduling process. In this case, depending on its complexity, a vehicular service can be divided into different tasks. An efficient task scheduling needs to orchestrate where and for how long such tasks will run, considering the available pools, the mobility of nodes, and the tasks deadline constraints. Thus, this article proposes an efficient VC task scheduler based on an approximation heuristic and resources prediction to select the best VC for each task, called MARINA. MARINA aims to analyze the behavior of vehicles that share their computational resources with the VC and make scheduling decisions based on the mobility (VC availability) of these vehicles. Simulation results under a realistic scenario demonstrate the efficiency of MARINA compared to existing state-of-the-art mechanisms in terms of the number of tasks scheduled, monetary cost, system latency, and Central Processing Unit (CPU) utilization.

Vehicular Edge Computing is a promising paradigm that provides cloud computing services closer to vehicular users. Vehicles and communication infrastructure can cooperatively provide vehicular services with low latency constraints through vehicular cloud formation and using these computational resources via task scheduling. An efficient task scheduler must decide which cloud will run the tasks, considering vehicular mobility and task requirements. This is important to minimize processing time and, consequently, monetary cost. However, the literature solutions do not consider these contextual aspects together, degrading the overall system efficiency. This work presents EFESTO, a task scheduling mechanism that considers contextual aspects in its decision process. The results show that, compared to state-of-the-art solutions, EFESTO can schedule more tasks while minimizing monetary cost and system latency.

Due to the increased computational capacity of Connected and Autonomous Vehicles (CAVs) and worries about transferring private information, it is becoming more and more appealing to store data locally and move network computing to the edge. This trend also extends to Machine Learning (ML) where Federated learning (FL) has emerged as an attractive solution for preserving privacy. Today, to evaluate the implemented vehicular FL mechanisms for ML training, researchers often disregard the impact of CAV mobility, network topology dynamics, or communication patterns, all of which have a large impact on the final system performance. To address this, this work presents FLEXE, an Open Source extension to Veins that offers researchers a simulation environment to run FL experiments in realistic scenarios. FLEXE combines the popular Veins framework with the OpenCV library. Using the example of traffic sign recognition, we demonstrate how FLEXE can support investigations of FL techniques in a vehicular environment.

Trajectory and traffic flow prediction will play an essential role in Intelligent Transportation Systems (ITS) to enable a whole new set of applications ranging from traffic management to infotainment applications. In this scenario, deep learning approaches such as Recurrent Neural Networks (RNN) and its variant Long Short Term Memory (LSTM) are excellent alternatives due to their ability to learn spatiotemporal dependencies. However, these neural networks tend to be over-complex and hard to design due to the broad set of hyper-parameters. We propose an automated framework to predict future trajectories and traffic flows in urban areas without human interventions. We employ Reinforcement Learning (RL) and Transfer Learning (TL) to generate high-performance LSTM predictors, which is referred as RL-LSTM. In addition, we introduce HERITOR (High ordE r tR affI c convoluTiOn R 1-lstm), a novel deep learning algorithm for traffic flow prediction. Specifically, HERITOR attempts to capture pure spatiotemporal features of urban traffic. The extracted features are fed into the RL-LSTM to realize a high performance LSTM for traffic flow prediction. We examine the proposed trajectory and traffic flow predictors on two real-world, large-scale datasets and observe consistent improvements of 15% - 25% over the state-of-the-art.

The use of contextual data to suggest distinct types of routes helps to understand new aspects of a city that may change the perception of drivers about routes. The impact of these aspects may differ from driver to driver requiring a way to change the suggestion according to the driver's point of view. Therefore, this paper presents an approach that identifies distinct situations in multiple types of contextual data and proposes a personalized and context-aware vehicle rerouting service called PONCHE. It considers common characteristics found in every dataset of spatiotemporal data to overcome the necessity of processing specific aspects of distinct data types. Regarding personalized service, each driver's profile is reflected into contextual data type weights considered by the system, i.e., the intensity he/she wants to avoid a contextual region. With that, a driver's profile may ignore a determined contextual data type. Performance evaluation results show that PONCHE identifies the best routes according to the weights given by a driver. It also improves the quality of contextual information obtained according to traffic, crime, and vehicle crashes. This study takes into consideration contextual data from Austin and Chicago in the USA, enabling comparison with two distinct cities.

Contextual data characterize distinct regions of the city, allowing them to differentiate them according to security, entertainment, services, among others. Using contextual data to suggest routes helps to understand new aspects of a city that can change users’ perceptions of different routes. The impact of each type of contextual data may vary according to the user’s profile, which is not taken into account in most of the systems proposed by the literature. Besides, it is necessary to consider the behavior of contextual data, which changes according to the type of data. To tackle the problems mentioned above, we propose a route suggestion system with space-time risk, called GIN. The system consists of three modules, namely: identification of contextual windows, context mapping, and route personalization. Moreover, we propose a strategy to decrease the number of route requests to improve system scalability. The results show that the system adapts to sensitive changes in user’s profiles. We obtained promising by using the behavior of contextual data to avoid unnecessary requests. This strategy allowed a reduction of up to 50% of requests made to the system.

Traffic congestion is a major problem in metropolitan areas, which inevitably leads to substantial social and economic impacts. In the Connected Autonomous Vehicles (CAVs) context, Intelligent Transportation System (ITS) addresses routing techniques for building an efficient transportation system in an urban environment. In order to improve traffic management, CAVs use real-time traffic data to disseminate faster routes for vehicles. Meanwhile, Cloud Computing is used to manage the traffic congestion situation, but it is not a suitable option for low-latency requirements of autonomous vehicles. Fog-based approaches dealing with traffic congestion found in the literature do not consider the use of caching for a routing scheme. Therefore, we propose a reliable caching mechanism for autonomous vehicle path planning based on Fog Computing, which is called ReCall. ReCall caches real-time traffic information from different regions to dynamically perform route recommendations. The results have shown that ReCall is able to reduce travel time and emissions.

Vehicular traffic re-routing is key to provide better vehicular mobility. However, considering just traffic-related information to recommend better routes for each vehicle is far from achieving the desired requirements of a good Traffic Management System, which intends to improve not only mobility but also driving experience and safety of drivers and passengers. Context-aware and multi-objective re-routing approaches will play an important role in traffic management. However, most of these approaches are deterministic and can not support the strict requirements of traffic management applications, since many vehicles potentially will take the same route, and, thus, degrade the overall traffic efficiency. In this work, we introduce Safe and Sound (SNS), a non-deterministic multi-objective re-routing approach for improving traffic efficiency and reduce public safety risks (based on criminal events) for drivers and passengers. SNS employs a hybrid architecture and a cooperative re-routing approach for improving system scalability and computation efforts. SNS uses a recurrent neural network to both predict future safety risks dynamics and enable a personalized re-routing in which each vehicle decides the risks it wants to avoid. Simulation results revealed that when compared to state-of-the-art approaches, SNS reduces the CPU time of the re-routing algorithm in approximately 99% and decreases the average safety risk for drivers and passengers in at least 30% while keeping efficient traffic mobility.

Recently, many cities are facing mobility and safety issues, commonly related to traffic congestion and the high number of city-wide criminal incidents. Several Intelligent Transportation Systems (ITS) were proposed to overcome mobility issues. Meanwhile, some safety-based systems were proposed to guide pedestrians and drivers toward safest paths. However, most of these systems address only a single issue. Hence, in order to avoid traffic congestion, an ITS may guide vehicles toward risky areas, while a safety-based system may guide them toward congested ones, focusing only on the safety of drivers and passengers. This paper introduces FnS (Faster and Safer), a hybrid ITS which employs accurate knowledge about traffic conditions and unsafety levels on roads for improving the safety of drivers and passengers at the same time it deals with traffic congestion. Simulation results under a realistic scenario have shown that FnS outperformed state-of-the-art approaches that deal with mobility or safety issues.

With the growth of large urban centers, some problems arise and solutions must be sought to contain them. In this context, traffic congestion is one such problem, where road infrastructure does not follow the high growth in the number of vehicles. Traffic Management Systems (TMS) arise to mitigate traffic-related problems, with automatic detection of slow roadways and vehicle rerouting to avoid such routes. Such applications are supported by the Vehicular Ad hoc NETworks (VANETs), where traffic information are disseminated between the vehicles or central server, and thus the better decisions about traffic management can be made. However, the data dissemination in VANETs is a challenging task, due to the short-range communication and high node mobility. Thus, this paper introduces a protocol for Data Dissemination based on Metrics of Complex Networks, called CRONOS. It provides data dissemination of traffic data with low overhead and high coverage. Simulation results show that CRONOS reduced the number of transmissions by 95%, the congestion time by 48.95%, and travel time by 18.11% for a TMS application.

Recently, many cities are facing challenging mobility and safety issues. The former is commonly related to traffic congestion, as a consequence of uncontrolled population growth and accelerated urbanization. The latter regards to elevated number of city-wide criminal incidents. Several Intelligent Transportation Systems (ITS) were proposed to overcome mobility issues; meanwhile, some safety- based systems were proposed to guide pedestrians and drivers toward safest paths. However, most of these systems tackle only one of the issues. Hence, an ITS can guide vehicles toward risky areas, in order to avoid traffic congestion, while a safety-based system can guide them toward congested roads, focusing on the safety of drivers and passengers. This paper introduces itsSAFE (Intelligent Transportation Systems for improving SAfety and traFfic Efficiency), an ITS which employs accurate knowledge about traffic conditions and unsafety levels on roads for improving the safety of drivers and passengers at the same time it deals with traffic congestion. Simulation results under a realistic scenario have shown that itsSAFE outperformed state-of-the-art approaches that deal with mobility or safety issues, by effectively dealing with traffic efficiency and safety.

Urban mobility has become one of the most challenging issue in urban centers. As a consequence, traffic congestion has become a daily problem. Several Advanced Traffic Management Systems (ATMS) have been proposed to improve overall traffic efficiency. However, these systems inefficiently exchange traffic information, which can lead to network overload. In order to overcome the mobility problem and improve the efficiency in dealing with vehicle traffic, this paper introduces a fully-distributed advanced traffic management system based on opportunistic content sharing, named PANDORA. Simulation results indicate that PANDORA outperforms the assessed solutions in various scenarios, considering different key requirements of ATMS.

Vehicular Ad-hoc NETworks (VANETs) are composed of moving vehicles with the ability to process, store, and communicate via wireless medium. VANETs promise a wide scope of services, such as, safety and security, traffic efficiency, and others. For instance, a VANET application can detect, control and reduce traffic congestion based on data that describes traffic patterns. However, disseminating data in VANET is a challenging task, due to its particular characteristics, i.e., heterogeneous density, short-range communication, and node mobility. Since, existing protocols for data dissemination do not effectively address the high overhead, in this paper, we proposed a Data Dissemination protocol Based on Centrality (DDBC) for urban scenarios. The simulation results show that DDBC protocol offers good efficiency in terms of delays and overhead, while achieve network coverage around 90%.

Congestion is a major problem in large cities. One of the main causes of congestion is the sudden increase of vehicle traffic during peak hours. Current solutions are based on perceiving road traffic conditions and re-routing vehicles to avoid the congested area. However, they do not consider the impact of these changes on near future traffic patterns. Hence, these approaches are unable to provide a long-term solution to the congestion problem, since when suggesting alternative routes they create new bottlenecks at roads closer to the congested one, thus just transferring the problem from one point to another. With this issue in mind, we propose an intelligent traffic system called CHIMERA, which improves the overall spatial utilization of a road network and also reduces the average vehicle travel costs by avoiding vehicles from getting stuck in traffic. Simulation results show that our proposal is more efficient in forecasting congestion and is able to re-route vehicles appropriately, performing a proper load balance of vehicular traffic.

In recent years, the number of vehicles has increased faster than the available infrastructure. Consequently, traffic congestion has become a daily problem affecting several aspects of modern society, including regional economic development. In this way, Traffic Management System (TMS) have been proposed to improve the traffic efficient and minimize traffic congestion problems. These systems rely on gather traffic-related data in a central entity to identify congestion and suggest alternative routes. However such approach adds load in communication channel depending on the traffic density. In this way, this paper introduces FASTER, a fully-distributed TMS to improve the overall vehicle traffic efficiency that does not overloads the communication channel, providing a suitable distributed solution. Simulation results indicate that our FASTER outperforms the assessed solutions in different scenarios and in different key requirements of TMS.

As an important component of Smart Cities, transportation system plays a critical role to address the sustainability and mobility of the society. One key concern is that the number of vehicles continuously increases faster than the available infrastructure, as well as the traffic congestion is a difficult issue to deal with. Several solutions to Intelligent Transportation Systems (ITS) have been proposed to identify congestion and re-route the vehicles afterwards. In this direction, this work introduces SPARTAN, a fully distributed ITS solution, which notifies drivers about congested areas through Vehicle-to-Vehicle communication and employs a real-time decision making mechanism used to reroute vehicles to avoid the congested areas. Simulation results show the effectiveness of SPARTAN in calculating new routes and disseminating them to vehicles that approaching a congestion area. As a consequence, SPARTAN reduces the travel time and the congestion time in urban scenarios when compared to existing approaches.

Traffic jams frustrate drivers and cost billions per year in time and fuel consumption. In order to avoid such problems, this paper presents an intelligent transportation system that collects real-time traffic information and is able to detect and manage traffic congestion based on this information. Simulation results show that the proposed protocol can reduce the average travel time, CO2 emission and fuel consumption. In particular, the average travel time was reduced in approximately 23%, the average fuel consumption in 9% and average CO2 emission in 10%.

Vehicular Ad-Hoc Networks (VANETs) are a specific type of moving networks in which the nodes are vehicles with processing, storage and wireless communication capacity. VANETs face a number of challenges in terms of data dissemination due to the volatile density of vehicles and frequent changes in the network topology induced by the high mobility of the vehicles and of short-range communications. The envisaged applications, as well as some inherent characteristics of the VANETs render the data dissemination an essential service and a challenging task in these networks. Many data dissemination protocols have been proposed in the literature, nevertheless, most of such protocols do not deal simultaneously with the problems of broadcast storm and synchronization problem. To face such problems, we propose a new data dissemination protocol in vehicular networks named ADD, which operates in highly dynamic highway environments. The ADD consists two mechanisms, broadcast suppression and delay desynchronization. ADD uses a preference zone to eliminate the broadcast storm problem and the delay desynchronization to eliminate the synchronization problem caused by 802.11p protocol. When compared with three known solutions SRD, Flooding and AID we show that our proposal for data dissemination executes it with higher efficiency than other protocols, exceeding them in different scenarios in all the undertaken evaluations.