High availability is a fundamental requirement in large-scale distributed systems, where replication strategies are central in keeping applications operational despite a bounded number of failures. State Machine Replication (SMR) is one of the most widely adopted approaches for implementing highly available, fault-tolerant services, as it increases uptime while ensuring strong consistency. In recent years, research on SMR has yielded numerous variations tailored to enhance resilience, performance, and scalability. In this paper, we revisit SMR from a new perspective by introducing Composing State Machine Replication (CSMR), a method that enables fault-tolerant service composition. By composing SMRs, we promote the reuse of existing services to construct more complex and reliable systems. This modular approach fosters loosely coupled, flexible architectures, contributing to the theoretical foundations of SMR and aligning with common development practices in cloud computing and microservices. We formally define CSMR and demonstrate how composition can be used to extend existing SMR specifications with new features. For example, CSMR allows the semantics of a service operation to be extended by enabling different state machine replicas to execute complementary steps of the same operation. Additionally, SMR composition facilitates sharding and state partitioning by assigning disjoint state variables to separate SMRs. Beyond formalization, the paper provides illustrative examples of CSMR and introduces a high-level CSMR architecture that highlights the essential components, their responsibilities, and their interactions in supporting the composition process. To further demonstrate practicability, we present an API for building CSMR systems that combines RPC-based communication with declarative configuration in YAML format.

Agriculture faces significant challenges from crop diseases, making early and accurate detection critical. Federated Learning (FL), an advancement in artificial intelligence (AI) and machine learning (ML), presents a promising solution by enabling collaborative model training on decentralized data without the need to share sensitive information. This article examines the application of FL in detecting plant diseases through image analysis, highlighting the role of cloud computing in addressing challenges related to data processing, storage, and model scalability. By leveraging decentralized data stored and processed in the cloud, FL develops robust models that not only improve detection accuracy but also generalize effectively to new data, promoting knowledge sharing while ensuring data privacy. The integration of cloud infrastructure enables FL to scale, providing resilience and productivity gains in agricultural practices. The results show that the proposed approach achieves a 99.71% accuracy using the VGG16 model after Federated Learning aggregation, while preserving data confidentiality, enhancing agricultural resilience, and benefiting from the scalability and flexibility offered by cloud computing.

Counting people in various urban spaces using artificial intelligence enables a wide range of smart city applications, enhancing governance and improving citizens' quality of life. However, the rapid expansion of edge computing for these applications raises concerns about the growing volume of electronic waste. To address this challenge, our previous work demonstrated the feasibility of repurposing confiscated illegal TV boxes as Internet of Things (IoT) edge devices for machine learning applications, specifically for people counting using images captured by cameras. Despite promising results, experiments in crowded scenarios revealed a high Mean Absolute Error (MAE). In this work, we propose a patching technique applied to YOLOv8 models to mitigate this limitation. By employing this technique, we successfully reduced the MAE from 8.77 to 3.77 using the nano version of YOLOv8, converted to TensorFlow Lite, on a custom dataset collected at the entrance of a university restaurant. This work presents an effective solution for resource-constrained devices and promotes a sustainable approach to repurposing hardware that would otherwise contribute to electronic waste.

Regarding privacy laws and digital globalization, understanding data regulation compliance and cross-jurisdictional challenges remains limited. To avoid administrative sanctions and protect user data, organizations and developers must bridge these gaps, navigating laws such as the General Data Protection Regulation (GDPR), the American Data Privacy and Protection Act (ADPPA), the General Data Protection Law (LGPD), and the Australian Privacy Act. This study focuses on creating a comprehensive compliance tool by investigating the similarities and nuances of these laws, as well as the challenges developers and organizations face in implementing Privacy by Design principles and ISO/IEC 29100 standards. Through a Systematic Literature Review (SLR) approach, topics of convergence and divergence among privacy laws and frameworks were pinpointed, as well as the challenges of implementing these laws in software. A survey was used to validate the challenges found in the SLR in the Brazilian context, in which most participants demonstrated a lack of knowledge regarding the LGPD. Lastly, we applied Framework Analysis to code and index key legislation points, allowing us to correlate them and develop a compliance-assistance tool. In the several contributions achieved, there is a deeper understanding of the privacy implications in a global context and its practical challenges, and also a practical guidance development, translating legal requirements into actions. Some limitations in this study lie in the interaction between selection and treatment in the survey, as participants' responses will not necessarily serve to generalize the challenges faced by all developers and organizations. In general, the contributions offer valuable theoretical and practical insights in the field of data privacy.

Companies and Internet Service Providers (ISPs) apply monitoring tools over network infrastructure, encompassing regular performance evaluations, with a primary focus on delivering crucial information about the current state of the network infrastructure and, consequently, the services running on it. However, these monitoring tools require ongoing development to handle more complex tasks, such as detecting performance issues. Within this context, this article proposes a mechanism for identifying high delays and communication links in the network that may cause these performance issues, using a temporally formulated Impact Score. This Score is based on data correlation techniques applied to information collected by monitoring tools. Experiments conducted with real data from the RNP Network indicate the efficiency of the proposal in identifying links impacting data communication, resulting in high end-to-end delays.

Providing knowledge about the characteristics of diverse cultural groups worldwide and identifying cultural similarities between their respective occupation regions can yield significant economic and social benefits. However, much of the existing research in this field relies on user behavior data, which may limit scalability and generalization due to the difficulty in obtaining such data. To address this, our work focuses on extracting venue data from Google Places and proposing a methodology based on the Scenes concept to enrich this dataset for generating cultural signatures of urban areas. This approach also considers the influence of different area sizes. Using Curitiba, Brazil, and Chicago, USA, as case studies, the results demonstrate that the proposed method can identify cultural similarities between regions while supporting an area-division strategy for analyzing cities across different countries. The findings show consistency, as evidenced by the segmentation of Curitiba and Chicago into culturally distinct clusters. This highlights the societal benefits of the proposal, such as location recommendations based on cultural criteria and real-time service validation.

When predicting the next geolocation of a stolen vehicle using external sensor data, such as speed radars, the challenge extends beyond the prediction itself to include determining the most suitable prediction architecture. While existing studies provide data that influence prediction performance, there is no consensus on the optimal architecture. Therefore, adopting a broader perspective to identify key criteria influencing the choice of architecture is essential. This study evaluates the shift in the optimal architecture depending on the length of the historical sequence and the format of geographic representation. The results reveal a shift in the optimal architecture, with the shift point being influenced by the type of geographic representation.

Data streams produced by mobile devices, such as smartphones, offer highly valuable sources of information to build ubiquitous services. Such data streams are generally uploaded and centralized to be processed by third parties, potentially exposing sensitive personal information. In this context, existing protection mechanisms, such as Location Privacy Protection Mechanisms (LPPMs), have been investigated. Alas, none of them have actually been implemented, nor deployed in real-life, in mobile devices to enforce user privacy at the edge. Moreover, the diversity of embedded sensors and the resulting data deluge makes it impractical to provision such services directly on mobiles, due to their constrained storage capacity, communication bandwidth and processing power. This article reports on the FLI technique, which leverages a piece-wise linear approximation technique to capture compact representations of data streams in mobile devices. Beyond the FLI storage layer, we introduce Divide & Stay, a new privacy preservation technique to execute Points of Interest (POIs) inference. Finally, we deploy both of them on Android and iOS as the INTACT framework, making a concrete step towards enforcing privacy and trust in ubiquitous computing systems.

We address the challenge of managing access to Content Delivery Networks (CDNs). In particular, we consider a scenario where users request tokens to access content, and one form of piracy consists in illegally sharing tokens. We focus on mitigating token misuse through performance analysis and statistical access pattern monitoring. Specifically, we examine how illegal token sharing impacts content delivery infrastructure and propose defining acceptable request limits to detect and block suspicious access patterns. Additionally, we introduce countermeasures against piracy, including selective quality degradation for users identified as engaging in illegal sharing, aiming to deter such behavior. Using queuing models, we quantify the impact of piracy on system performance across different scenarios. To validate our model, we perform statistical tests that compare real CDN traffic patterns with the expected request intervals in our proposed framework. These measures—defining access thresholds, quality degradation for unauthorized use, and statistical alignment checks—enhance CDN access management, preserving infrastructure integrity and the legitimate user experience while reducing operational costs.

Road safety remains a global challenge, especially in scenarios where behavioral and environmental factors heavily influence drivers' decision-making. Machine learning models play a crucial role in enhancing safety and informed decision-making by learning effective actions based on traffic conditions. However, training these models requires access to user data, which can compromise drivers' privacy and expose sensitive information. To address this issue, this study proposes a solution for generating synthetic driving condition data using a Federated Learning approach combined with Generative Adversarial Networks (GAN). This method enables model training across multiple federated learning clients, preserving data privacy by avoiding direct data sharing. By leveraging the Harmony dataset, similarity metrics such as Euclidean Distance and KL-Divergence were integrated into the GAN loss function to improve the quality of the generated synthetic data. The results demonstrate that the proposed approach successfully generates realistic driving condition data, supporting centralized model training while maintaining user privacy, showcasing its potential in privacy-conscious road safety applications.