O monitoramento de tráfego de rede é essencial para compreender o comportamento da infraestrutura e avaliar a integridade de seus componentes. O aprendizado federado tem se destacado como uma abordagem promissora para sistemas de defesa baseados nesse monitoramento, permitindo o treinamento distribuído de modelos sem compartilhamento direto de dados. No entanto, métodos tradicionais assumem um ambiente federado composto apenas por clientes honestos, ignorando a possibilidade de ataques de envenenamento de rótulos (label poisoning). Este trabalho propõe um novo arcabouço de aprendizado federado robusto contra ataques de rede, com foco na mitigação de clientes maliciosos. Nossa abordagem emprega técnicas de Redes Siamesas para quantificar a aderência dos dados e ajustar dinamicamente a ponderação das contribuições de cada cliente, fortalecendo a resiliência do modelo contra manipulações adversárias. Os resultados mostram que nossa estratégia não apenas melhora a detecção de ataques, mas também reduz significativamente o impacto de envenenamento de rótulos no aprendizado federado.

In the age of the Internet of Things and the Internet of Vehicles, mobile entities generate massive and unrestricted trajectory data that pose privacy concerns, such as linking attacks. Mix-zones have been used to anonymize citizens' identities to address privacy issues. However, the mix-zones depend on factors, such defining fair privacy levels, that impact their performance, privacy level, and anonymization quality. This work proposes the Dynamic-k Mix-zone (Dk-Mix), a dynamic mix-zone that tunes the privacy level over time in an online model with linear complexity, according to vehicles' traffic fluctuations, to achieve higher anonymization. We evaluated our approach on real and synthetic datasets, comparing it against two privacy-level prediction mechanisms and classical mix-zones regarding coverage, privacy metrics, and Anonymization Quality (AQ). Results showed that Dk-Mix achieves more accurate privacy-level estimation, with efficacy, anonymization rate, and AQ comparable to the best classical mix-zones. Moreover, it maximized privacy over time while minimizing trajectory re-identification up to 30.97% and 40.92% for high and low traffic, outperforming traditional mix-zones.

Due to the environmental impact caused by greenhouse gas emissions, solving problems aimed at increasing the usage of electric vehicles became important. Personal Electric Vehicles are being highly adopted by society in order to reduce emissions. However, a prominent part of air pollution is provided by heavy-duty vehicles, such as trucks, and its electrification is challenging because of the lack of government policies and charging infrastructure. In light of this, electric charge stations should be located considering the truck drivers’ route to increase its adoption. Therefore, this study proposes two hexagonal discrete covering models, a Hexagonal P-Median (HPMP) and Hexagonal Capacitated Location Set Covering (HCLSCP), enhancing the space complexity of classical discrete models to cover the Brazilian truck drivers’ route. Furthermore, we compare the novel hexagonal models to a greedy method using a spatio-temporal simulation. We consider the infrastructure limitations with capacity constraints and waiting time in recharging queues with real-world data comprising locations of 3,086 drivers. The results show a trade-off between infrastructure cost, coverage demand, and queuing performance. The HPMP is ideal for covering demand, while the greedy method minimizes infrastructure cost, and HCLSCP outperforms the other models in queuing management.

Devido às suas restrições computacionais e configurações incorretas, os dispositivos da Internet das Coisas (IoT) são alvos fáceis de diversos ataques. Neste trabalho, propomos uma nova abordagem baseada nas transformações de padrões ordinais para a identificação remota de Sistemas Operacionais (SOs), uma etapa fundamental para identificar possíveis vulnerabilidades nesses dispositivos. Para isso, é analisado o comportamento dinâmico dos números iniciais de sequência (ISN) do cabeçalho do TCP. Verificamos a capacidade do método na detecção de similaridades e diferenças entre SOs clássicos e modernos, comparando com dispositivos da IoT. Os experimentos comprovam sua eficácia em reconhecer os SOs pelos seus diferentes padrões de geração de ISN, superando ferramentas consolidadas como o Nmap, bem como sendo capaz de classificá-los com uma acurácia de 100% em certos casos.

Researchers have studied how to improve Federated Learning (FL) in various areas, such as statistical and system heterogeneity, communication cost, and privacy. So far, most of the proposed solutions are either very tied to the application context or complex to be broadly reproduced in real-life applications involving humans. Developing modular solutions that can be leveraged by the vast majority of FL structures and are independent of the application people use is the new research direction opened by this paper. In this work, we propose a plugin (named FedPredict) to address three problems simultaneously: data heterogeneity, low performance of new/untrained and/or outdated clients, and communication cost. We do so mainly by combining global and local parameters (which brings generalization and personalization) in the inference step while adapting layer selection and matrix factorization techniques to reduce the downlink communication cost (server to client). Due to its simplicity, it can be applied to federated learning of different number of topologies. Results show that adding the proposed plugin to a given FL solution can significantly reduce the downlink communication cost by up to 83.3% and improve accuracy by up to 304% compared to the original solution.

A análise das práticas de condução veicular tem ganhado relevância crescente devido à sua ampla aplicabilidade em sistemas inteligentes de transporte, especialmente no que se refere à eficiência energética no setor automotivo. Nesse cenário, o desenvolvimento de métodos eficazes para rotular e classificar estilos de condução é essencial. Este estudo propõe uma metodologia para atribuição de rótulos a estilos de condução veicular, utilizando dados de telemetria automotiva coletados via barramento Controller Area Network (CAN). A abordagem integra o agrupamento não supervisionado K-Means com múltiplas estratégias de classificação, incluindo um modelo de lógica fuzzy, um processo determinístico e um método baseado em indicadores estatísticos. Os resultados demonstram que a combinação dessas técnicas proporciona uma interpretação mais precisa e robusta dos estilos de condução. Essa validação cruzada entre as técnicas reforça o potencial de abordagens não supervisionadas na identificação de perfis de condutores, com aplicações relevantes para o setor automotivo, especialmente aquelas voltadas à sustentabilidade e ao uso mais racional de recursos.

The Internet of Drones (IoD) is an emerging technology that will enable a new era of drone services and applications. However, many barriers and challenges remain until it is possible to control a complex IoD network. The scientific community is still discussing, studying, and investigating the best way to implement this network to become the IoD viable, reliable, and efficient. Furthermore, the principles that guide terrestrial wireless networks and even traditional Unnamed Aerial Vehicles (UAV) networks do not apply to IoD mainly because it allows distinct drones performing different applications to share the airspace. This thesis aims to provide procedures and discussions that can guide future development to overcoming barriers related to fundamental problems in IoD, such as communication and mobility.

Simulation is the most commonly used method for evaluating protocols and algorithms within vehicular networks (VANETs). Typically, simulation tools employ mobility traces to reconstruct the network topology, which relies on existing interactions among mobile nodes. However, the quality of these traces, particularly their spatial and temporal resolution, plays a critical role in accurately shaping the network topology. Consequently, the validity of simulation outcomes heavily depends on the mobility model’s ability to mirror the real network topology accurately. We demonstrate that actual bus mobility traces exhibit gaps, which lead to outcomes that fail to represent reality accurately. In this study, we introduce a method to address these gaps, resulting in enhanced characteristics that contribute to more trustworthy simulation outcomes. Moreover, we present evaluation results comparing the communication metrics of both the original and adjusted traces. These findings indicate that the gaps create network topologies that deviate from the actual scenario, diminishing the credibility of the evaluation results. Although mobility data for vehicular networks is beneficial, it necessitates a process of quality enhancement.

The 6G wireless networks are already being studied and their main novelties are beginning to emerge. The 6G network promises 100% network coverage in every area, including wilderness areas such as the ocean, desert, or remote areas, as it will be a fully 3D network, i.e., it will integrate space, air, and terrestrial networks. Also, with all these types of devices, the complexity of 6G networks increases, and Artificial Intelligence (AI) techniques will be mandatory for all the 6G’s potential data and parameters. Hence, a Space-Air-Ground Integrated Network (SAGIN) is a potential alternative to enable all the network coverage as drones and satellites can cover large areas and fill possible network coverage gaps. The 6G network coverage also includes post-disaster regions with no network infrastructure and no network coverage, making SAGIN very suitable for them. Thus, in this work, we expand CAIN, a post-disaster 6G routing protocol, so it can also be used in SAGIN scenarios, increasing even more network coverage. With drones as relays, SAGIN-CAIN routes the CAIN’s Cluster Head’s messages until it reaches a satellite.

Detecting transportation modes' usability in spatiotemporal urban trajectories can provide valuable insights into the mobility preferences of urban populations, helping epidemic prevention and urban quality-of-life improvement. With this goal, we introduce POPAyI, a strategy that bases its design on the Ordinal Pattern (OP) transformation applied to mobility-related time series. POPAyI can quantify time-series dynamics with a low-complex cost, muscling time series' characteristics without the need for high computational and methodological complexities as the current Machine Learning (ML) and Deep Learning (DL) literature. POPAyI uses polar representation and captures amplitude information in time series, bringing the multivariate capability to the standard 1D OP transformation. Our experiments show that POPAyI: (i) perfectly adapts to multi-dimensional mobility time series and natural non-linear mobility behavior. (ii) presents consistent detection results in any considered number of transportation mode's classes with efficiency in terms of storage and computation complexity, using fewer features than ML approaches and computational resources than DL methods, e.g., reaching 10000 fewer parameters than a lightweight DL approach while increasing by 3% the F1-score.

Intelligent Transportation Systems (ITS) faces significant challenges in achieving its goal of sustainable and efficient transportation. These challenges include real-time data processing bottlenecks caused by high communication latency and security vulnerabilities related to centralized data storage. We propose a novel architecture that leverages Edge Computing and Distributed Ledger Technology (DLT) to address these concerns. Edge computing pushes cloud services, such as vehicles and roadside units, closer to the data source. This strategy reduces latency and network congestion. DLT provides a secure, decentralized platform for storing and sharing ITS data. Its tamper-proof nature ensures data integrity and prevents unauthorized access. Our architecture utilizes these technologies to create a decentralized platform for ITS data management. This platform facilitates secure processing, storage, and data exchange from various sources in the transportation network. This paper delves deeper into the architecture, explaining its essential components and functionalities. Additionally, we explore its potential applications and benefits for ITS. We describe a case study focusing on a data marketplace system for connected vehicles to assess the architecture’s effectiveness. The simulation results show an average latency reduction of 83.35% for publishing and 87.57% for purchasing datasets compared to the cloud architecture. Additionally, transaction processing speed improved by 18.73% and network usage decreased by 96.67%. The proposed architecture also achieves up to 99.61% reduction in mining centralization.

O aprendizado federado (FL) surgiu como uma técnica onde diversos dispositivos (também chamados de clientes) podem aprender de forma colaborativa a partir da orquestração de um servidor central, proporcionando escalabilidade, privacidade e baixo custo de comunicação. A maioria das pesquisas sobre este tema apresenta propostas para a etapa do treinamento de modelos no aprendizado federado, para endereçar diversos problemas como a heterogeneidade estatística de dados, o que muitas vezes representa aumento de custos (e.g., computacional, armazenamento e comunicação). No entanto, recentemente foi proposta a solução FedPredict, um plugin que opera na etapa de predição do aprendizado federado, que quando adicionado pode melhorar significativamente o desempenho de diversas soluções tradicionais em cenários de heterogeneidade de dados, sem requerer qualquer modificação na sua estrutura original ou adição de treinamento. Nesta direção, este trabalho apresenta experimentos sobre uma nova descoberta: quanto mais heterogêneos são os dados, menos treinamento é necessário quando o FedPredict é adicionado, tornando o processo de aprendizado altamente eficiente.

The Internet of Drones (IoD) emerged as a novel mobile network paradigm. IoD is a unique environment with particular characteristics that differ from traditional ones (e.g., drones’ mobility and the fast network topology change), demanding compliance with security and privacy requirements. Likewise, IoD can suffer from novel drone-centered threats. The existent protection mechanisms (PMs) may not be adequate for the IoD environment since they may not embrace the IoD characteristics, also facing new threats. Therefore, the main goal of this dissertation is to study the design of PMs for the IoD, considering its particular characteristics. This study reveals a need to enhance current PMs to meet the IoD characteristics since they can not offer the same protection level. Our contributions advance the state-of-the-art on four fronts: new guidelines for IoD security and privacy field; novel location privacy PMs; novel anti-jamming PMs; and new strategies for automatic drone detection.

In Federated Learning (FL), personalization-based solutions have emerged to improve clients’ performance, considering the statistical heterogeneity of local datasets. However, these methods are designed for a static environment and the previously learned model becomes obsolete as the local data distribution changes over time. This problem, known as concept drift, is widespread in several scenarios (e.g., change in user habits, different characteristics of visited geolocations, and seasonality effects, among others) but needs to be addressed by most solutions in the literature. In this work, we present FedPredictDynamic, a plugin that allows FL solutions to support statistically heterogeneous stationary and non-stationary local data. The proposed method is a lightweight and reproducible modular plugin and can be added to various FL solutions. Unlike state-of-the-art concept drift techniques, it can rapidly adapt clients to the new data context in the prediction stage without requiring additional training. Results show that when context changes, FedPredict-Dynamic can achieve accuracy improvements of up to 195% compared to concept drift-aware solutions and 210.7% compared to traditional FL methods.

Unmanned aerial vehicles (UAVs)-also known as drones or Unmanned Aircraft-have found diverse applications in various fields owing to their significant advantages, including fast mobility and rapid deployment. UAVs are crucial in aerial networks, providing increased coverage and on-demand connectivity as mobile nodes. In recent years, UAVs have made room to leverage the Internet of Things (IoT) to the sky, enhancing air-to-ground communication and pointing toward the next generation of UAV networks. As this expansion is still in its early stages, there are several aerial network terminologies, each with similarities and differences, depending on the deployment domain and the services they offer. However, studies have yet to discuss these different terminologies consistently. Key aspects have yet to be thoroughly explored, such as the anticipated requirements for deploying these networks and how they relate to the various terminologies. This work systematically analyzes the existing terminologies of UAV networks, considering their requirements and applications, shedding light on their intersections and differences. Furthermore, we present the demands for the next generation of UAV networks and discuss how they impact the design of UAV-related applications, aiding in the design of new protocols, tools, and technologies for both industry and academia. Lastly, we highlight the emerging trends and challenges associated with deploying and integrating these networks.

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.

A elaboração de soluções que viabilizem uso de meios de transporte com energia elétrica tornou-se importante, devido aos impactos ambientais causados pelos gases emitidos por queima de combustíveis fósseis. No entanto, para que esse tipo de veículo seja adotado, é preciso investir na infraestrutura rodoviária, tal como pontos de recarga elétrica. Este trabalho apresenta o Hexagonal P-Median, um modelo de alocação de pontos de recarga que atende às trajetórias dos caminhoneiros brasileiros. O modelo proposto foi comparado com um algoritmo guloso e um modelo de cobertura de conjuntos por meio de uma simulação com dados reais de 44,5 milhões de registros de localização de 3,086 motoristas. O modelo proposto apresenta, aproximadamente, 230% e 276% a mais de cobertura que o algoritmo guloso e o modelo de cobertura de conjuntos, respectivamente, considerando o cenário de 10 km de desvio.

This comprehensive survey explores the connection between the Internet of Drones (IoD) and Urban Computing (UC). Drones offer substantial benefits across diverse urban applications encompassing public safety, transportation management, delivery services, and emergency response, among others. To ensure the seamless operation of these applications, efficient airspace management is imperative. Consequently, an essential foundation is to establish a communication infrastructure coupled with an accompanying supportive system that caters to drones, control bases, service providers, and the Internet. Termed the IoD, this framework facilitates drones’ control management, synchronization, coordination and monitoring of drones to attain their objectives effectively. Moreover, IoD applications are poised to propel societal advancement by aligning with the goals of UC. UC is an interdisciplinary domain to enhances human well-being by harnessing data from various sources, including the Global Positioning System (GPS), cameras, proximity sensors, and social networks. This survey thoroughly examines and analyzes multifaceted aspects underpinning how the IoD can significantly benefit UC. It encompasses applications, their requisites, design protocols specific to IoD, and the challenges inherent within this evolving realm. To the best of our knowledge, our work is one of the first studies to directly relate and show how the IoD can contribute to all UC-applications categories.

There is a shortage of mobility datasets - real or synthetic - available in the literature, limiting the development of new research. This generates demand for newer large-scale datasets, considering time, space, and population size. This work introduces a novel, two-step approach to generating large-scale mobility data considering the surrounding context, such as traffic. We validate our approach using two real mobility datasets, resulting in an enriched, large-scale dataset with more than 1 million origin-destination trips, which we share with the community to enable new research opportunities.

Researchers in Vehicular Edge Computing are witnessing a continuous search for a solution to the problem of how to best allocate computational resources to fulfill service requests from road vehicles efficiently. This problem combines several of the most difficult challenges associated with Intelligent Transportation Systems (ITSs), such as limited computational resources, dynamic vehicular network topology, high vehicle mobility, and long task execution times. These challenges represent significant barriers to the success of ITSs, severely impacting user experience and use of the service. Among alternatives, bioinspired algorithms have been used to support the complex decision-making associated with resource optimization due to their perceived success in simulating various natural behaviors and dealing with complex environments. However, to our knowledge, a comprehensive demonstration of their suitability and performance was never made when faced with the mentioned challenges. To fill this gap, we comprehensively investigate how the most prominent bio-inspired algorithms perform in challenging scenarios in Vehicular Edge Computing and compare them with other widely adopted alternatives. Our results show that bioinspired algorithms are both suitable and superior in efficiency, fulfilling a higher number of tasks.