Bruno Yuji Lino Kimura

A detecção dinâmica de aplicações móveis maliciosas requer rastros de execução amplos, recentes e representativos. Como a maioria das abordagens utiliza técnicas de aprendizado de máquina para essa detecção, há demanda por centralizar grandes volumes de dados para o treinamento dos modelos. Porém, essa centralização é dificultada por diversos desafios. A coleta de sequências de chamadas de sistema requer instrumentação custosa das aplicações, as diferenças entre dispositivos e versões do sistema geram distribuições heterogêneas de dados, e a concentração de dados compromete a privacidade e consome recursos críticos em dispositivos móveis e de borda. Assim, este trabalho propõe um mecanismo de detecção distribuído baseado em Aprendizado Federado, que permite treinar modelos localmente e compartilhar apenas suas atualizações com um servidor de agregação. O método alcança 0,900 de acurácia, 0,896 de F1-macro e 0,922 de AUPRC, além de reduzir significativamente o tráfego de dados em comparação com abordagens centralizadas, conciliando a efetividade das abordagens de aprendizado de máquina com as restrições de privacidade e recursos dos dispositivos.

A Computação de Borda Veicular (VEC) surge como um paradigma promissor para atender à crescente demanda por processamento de baixa latência em aplicações veiculares, impulsionada pelo aumento contínuo de veículos conectados e pela geração massiva de dados. No entanto, a natureza dinâmica e as restrições de recursos desse ambiente impõem desafios significativos para o escalonamento eficiente de tarefas computacionais. Neste trabalho, propõe-se o ORION, um escalonador baseado em otimização multiobjetivo que utiliza o algoritmo NSGA-II para equilibrar objetivos conflitantes: maximizar o número de tarefas concluídas dentro dos prazos, minimizar o custo monetário e reduzir a latência do sistema. Os resultados demonstram que o ORION supera outras soluções de escalonamento de tarefas em VEC, especialmente em cenários de alta demanda por recursos, mantendo altas taxas de escalonamento, reduzindo custos e alcançando níveis aceitáveis de latência.

The promises of multipath transport are to aggregate bandwidth, improve resource utilisation and enhance reliability. In this article, we demonstrate that the way multipath coupled congestion control is defined today leads to a suboptimal resource utilisation when network paths are disjoint, i.e., they do not share a bottleneck link. With growing interest in standardising Multipath QUIC (MPQUIC), we have implemented the practical Shared Bottleneck Detection (SBD) algorithm from RFC8382 in MPQUIC (MPQUIC-SBD). Through extensive experiments, we evaluate MPQUIC-SBD in the context of video streaming with various Adaptive Bitrate (ABR) algorithms, addressing both ABR classes of rule- and learning-based solutions. We demonstrate that MPQUIC-SBD accurately detects shared bottlenecks over 90% of the time, depending on the ABR algorithm, as the size of the video segments increases. In non-shared bottleneck scenarios, when MPQUIC-SBD detects that its QUIC subflows do not share the same network resources, it decouples their congestion windows accordingly, enabling video throughput gains of up to 37% compared to MPQUIC. These gains translate directly into improved video quality metrics, including higher bitrate, better resolution and reduced buffering, resulting in an enhanced quality of experience for users.

O crescimento de dispositivos conectados a Internet tem gerado grande volume de dados e, cada vez mais, impulsionado o uso de Inteligência Artificial (IA). Porém, abordagens centralizadas de aprendizado levantam preocupações relativas à privacidade dos usuários. O Aprendizado Federado (FL) desponta como alternativa distribuída, pois evita o compartilhamento de dados brutos dos usuários e atende legislações de privacidade. No entanto, a comunicação frequente entre dispositivos e servidor no FL gera alto consumo de banda e energia. Para redes com recursos limitados, soluções que mitiguem esse problema são cruciais. Nesse contexto, considerando a possibilidade de redução do tráfego no FL, este trabalho investiga técnicas de compressão de modelos, visando equilibrar qualidade do modelo e custo de comunicação. Os resultados indicam que as técnicas de compressão reduzem efetivamente o volume de dados transmitidos, sem degradar o desempenho do modelo, mesmo em cenários com dados desbalanceados.

Augmented and Mixed Reality (AR/MR) technologies enhance the human perception of the world by combining virtual and real environments. With the increase of mobile devices and the advent of 5G, this technology has the potential to become part of people’s life. This article aims to evaluate the impact of 5G and beyond mobile networks in the future of AR/MR. To attend to this objective, we surveyed four digital libraries to identify articles and reviews concerning AR/MR use based on mobile networks. The results describe the state-of-the-art of mobile AR/MR applications and the benefits and challenges of the technology. Finally, after the review, we propose a roadmap concerning AR/MR hardware and software development to run applications supported by future mobile networks.

The promises of multipath transport is to aggregate bandwidth and improve resource utilisation and reliability. We demonstrate in this paper that the way multipath coupled congestion control is defined today RFC6359 leads to a sub-optimal resource utilisation when network paths are mainly disjoint, i.e., they do not share a bottleneck. With growing interest to standardise Multipath QUIC (MPQUIC), we implement the practical shared bottleneck detection (SBD) algorithm from RFC8382 in MPQUIC, namely MPQUIC-SBD. We evaluate MPQUIC-SBD through extensive emulation experiments in the context of video streaming. We show that MPQUIC-SBD is able to correctly detect shared bottlenecks over 90% of the time as the video segments' size increase depending on the Adaptive Bitrate (ABR) algorithm. In non-shared bottleneck scenarios, MPQUIC-SBD results in video throughput gains of more than 13% compared to MPQUIC, which directly translates into better video quality metrics.

Weather events put human lives at risk mostly when people might reside in areas susceptible to natural disasters. Weather monitoring is a pivotal task that is accomplished in vulnerable areas with the support of reliable weather stations. Such stations are front-end equipment typically mounted on a fixed mast structure with a set of digital and magnetic weather sensors connected to a datalogger. While remote sensing from a number of stations is paramount, the cost of professional weather instruments is extremely high. This imposes a challenge for large-scale deployment and maintenance of weather stations for broad natural disaster monitoring. To address this problem, in this paper, we validate the hypothesis that a Low-Cost Automatic Weather Station system (LCAWS) entirely developed from commercial-off-the-shelf and open-source IoT technologies is able to provide data as reliable as a Professional Weather Station (PWS) of reference for natural disaster monitoring. To achieve data reliability, we propose an intelligent sensor calibration method to correct weather parameters. From the experimental results of a 30-day uninterrupted observation period, we show that the results of the calibrated LCAWS sensors have no statistically significant differences with the PWS's results. Together with The Brazilian National Center for Monitoring and Early Warning of Natural Disasters (Cemaden), LCAWS has opened new opportunities towards reducing maintenance cost of its weather observational network.

Interpath contention is a phenomenon experienced in a MultiPath TCP (MPTCP) connection when its subflows dispute resources of shared bottlenecks in end-to-end paths. Although solutions have been proposed to improve MPTCP performance in different applications, the impact of interpath contention on the multipath performance is little understood. In this paper, we evaluated such phenomenon experimentally in disjoint paths–an ordinary multipath scenario where subflows dispute bottlenecks of paths physically disjointed in a connection. Under several path conditions determined from emulations of capacity, loss, and delay of bottlenecks, we analyzed the influence of MPTCP mechanisms such as packet scheduling, congestion control, and subflow management. Differently from other studies, we observed that the very first influence was caused by the current subflow manager, full-mesh , with dichotomous impact on the multipath performance when establishing several subflows per disjoint path. Experimental results showed that contention among subflows can lead to positive (goodput improvement) or negative (goodput degradation) impacts according to the bottleneck conditions. In certain conditions, simply establishing subflows in single-mesh , with at most one subflow per disjoint path, could avoid interpath contention while improving goodput significantly, by doubling the performance of full-mesh under different conservative congestion controls.