Training in supervised machine learning is based on the availability of datasets; however, medical datasets must comply with stringent privacy regulations. Generative Adversarial Networks (GANs) are a relevant alternative to solve the limitation of small medical datasets due to their ability to generate additional data with desired features. A significant drawback of these models is that they may produce unrealistic, blurred, or insufficiently diverse images. This paper proposes a data augmentation technique using GANs to create synthetic Magnetic Resonance Imaging (MRI) of four stages of Alzheimer's Disease (AD): non-demented, very mild demented, mild demented, and moderate demented. We designed a GAN based on the Pix2Pix model, which learns the features of each AD stage. Generated images are evaluated by multistage Convolutional Neural Network (CNN) models, greyscale histograms of the distribution of pixel intensities, and brain mass measurements on binarized images. The results indicate that AD synthetic MRI effectively captures disease patterns, demonstrating the potential of GANs to improve training and diagnosis of neurodegenerative diseases.

This paper presents the design, development, and evaluation of PuzzleMesh, an agnostic service mesh composition model to process large volumes of data in edge-fog-cloud environments. This model is based on a puzzle metaphor where pieces, puzzles, and metapuzzles represent self-contained autonomous and reusable software artifacts encapsulated into containers and published as microservices. Apiecerepresents the integration of apps with I/O interfaces (loops/sockets), parallel processing, and management software. Apuzzlerepresents a processing structure (e.g., workflows) built coupling pieces through loops and sockets. Puzzles integrate structures with a microservice architecture, implicit continuous dataflows, and transparent data exchange management software. Ametapuzzlerepresents a recursive assemble of puzzles. A mesh represents a pool of pieces, puzzles, and metapuzzles available for designers to choose artifacts to build services. A prototype developed using PuzzleMesh model was evaluated through case studies about the automatic construction of processing services for the acquisition, pre-processing, manufacturing, preserving, and visualizing of satellite imagery. A qualitative comparison revealed that PuzzleMesh provides a flexible way to build reusable and portable services and to improve the usability of the services. The case study also revealed that PuzzleMesh yielded better performance results than other state-of-the-art tools.

The protection of data processing is emerging as an essential aspect of data analytics, machine learning, delegation of computation, Internet of Things, medical and financial analysis, smart cities, genomics, non-disclosure searching, among others. Often, they use sensitive information that cannot be protected by traditional cryptosystems. Homomorphic Encryption (HE) schemes and secure Multi-Party Computation (MPC) are considered suitable solutions for privacy protection. In this paper, we propose and analyze the performance of three homomorphic Logistic Regression (LR) models with Gradient Descent (GD) algorithms based on the Residue Number System (RNS). We compare their performance with four traditional non-homomorphic versions, one homomorphic algorithm based on RNS with Batch GD, and two state-of-the-art homomorphic algorithms. To validate our approach, we consider six public datasets of different medicine domains (diabetes, cancer, drugs, etc.) and genomics. We use a 5-fold cross-validation technique for a fair comparison in terms of the solution quality and training time. The results show that propose homomorphic solutions have similar accuracy with non-homomorphic algorithms, increased classification performance, and decreased training time compared with the state-of-the-art HE algorithms.

Abstract Classical machine learning modeling demands considerable computing power for internal calculations and training with big data in a reasonable amount of time. In recent years, clouds provide services to facilitate this process, but it introduces new security threats of data breaches. Modern encryption techniques ensure security and are considered as the best option to protect stored data and data in transit from an unauthorized third-party. However, a decryption process is necessary when the data must be processed or analyzed, falling into the initial problem of data vulnerability. Fully Homomorphic Encryption (FHE) is considered the holy grail of cryptography. It allows a non-trustworthy third-party resource to process encrypted information without disclosing confidential data. In this paper, we analyze the fundamental concepts of FHE, practical implementations, state-of-the-art approaches, limitations, advantages, disadvantages, potential applications, and development tools focusing on neural networks. In recent years, FHE development demonstrates remarkable progress. However, current literature in the homomorphic neural networks is almost exclusively addressed by practitioners looking for suitable implementations. It still lacks comprehensive and more thorough reviews. We focus on the privacy-preserving homomorphic encryption cryptosystems targeted at neural networks identifying current solutions, open issues, challenges, opportunities, and potential research directions.

Internet-of-Things (IoT) environment has a dynamic nature with high risks of confidentiality, integrity, and availability violations. The loss of information, denial of access, information leakage, collusion, technical failures, and data security breaches are difficult to predict and anticipate in advance. These types of nonstationarity are one of the main issues in the design of the reliable IoT infrastructure capable of mitigating their consequences. It is not sufficient to propose solutions for a given scenario, but mechanisms to adapt the current solution to changes in the environment. In this article, we present a multicloud storage architecture called WA-MRC-RRNS that combines the weighted access scheme, threshold secret sharing, and redundant residue number system with multiple failure detection/recovery mechanisms and homomorphic ciphers. We provide a theoretical analysis of the probability of information loss, data redundancy, speed of encoding/decoding, and show how to dynamically configure parameters to cope with different objective preferences, workloads, and cloud properties. We propose a multiobjective optimization mechanism to adjust redundancy, encryption-decryption speed, and data loss probability. Comprehensive experimental analysis with real data shows that our approach provides a secure way to mitigate the uncertainty of the use of untrusted and not reliable IoT infrastructure.

The numerical simulation of complex astrophysical problems requires high-performance computing due to the large size of the problems and variety of simulated physical processes. In this paper, we present a new framework for the numerical simulation of astrophysical fluid dynamics. It is based on the mechanisms of combining distributed and parallel computing techniques, advanced vectorization for KNL, and Skylake-SP CPU architectures. Our new HydroBox3D framework uses large 3D meshes to solve problems such as the dynamics of stars or galaxies. In our framework, we use computational nodes with a large amount of memory (RAM or Intel Optane in memory mode) for mesh processing and typical computational nodes for the numerical simulation of astrophysical problems. We use MPI both for send/receive operations between computational nodes and for sending processed data for calculations from data nodes. For optimization of calculations, memory, and CPU usage, we use data vectorization, FMA3, and AVX-512 instructions for Intel Xeon Phi 72XX and Intel Xeon Scalable processors. Benchmark results on different CPU and MIC devices show the effectiveness of the proposed solution.

The energy consumption of large data centers has been increasing for the last decades and currently is a major concern for economic and environmental reasons. Accurate scheduling of the data center operation and use of renewable energy sources present themselves as promising solutions for this problem. In this paper we study the problem of scheduling workflows of tasks in distributed heterogeneous data centers which are partially powered by renewable energy sources. This problem takes into account quality of service, infrastructure usage, and power consumption of machines and cooling devices. We propose a mathematical model for accurate scheduling solutions.

This article addresses a virtual machine (VM) allocation problem that appears in a novel business model for cloud computing. In this model, a cloud service broker owns a number of cloud reserved instances that outsources to its customers as cheap as on-demand VMs. The objective of the broker is to efficiently manage its reserved resources to maximize its revenue. We enhance the previous definition of the problem by considering more realistic parameters: geographical localization of resources and users, different types of applications, and data transfer costs. We propose a set of heuristics to solve the optimization problem of maximizing the cloud provider profit while offering appropriate Quality-of-Service to the users. The experimental analysis is performed over different scenarios using real data from cloud providers.

Frequency domain sensors (FDS) are important elements in control, data acquisition and monitoring systems. Such sensors have some outstanding characteristics like output of quasi-digital signals, high sensitivity, high resolution, wide dynamic range, anti-interference capacity and good stability. A FDS converts a physical variable (measurand) into a frequency domain ouput. When the measurand changes, the output has a proportional frequency shift. In systems that use FDS, measuring the frequency shift is desirable. Accuracy of most frequency measurement techniques is limited by measurement time, and if more precision is required, longer times for measuring are needed. In this work, a novel approach using the rational approximations principle for measuring frequency shift in the output of a FDS is introduced. Also algorithms for simulating the mathematical model of frequency measurement process are proposed, and resolution of measurement is improved by analyzing the data obtained.

The review of the requirements of different cloud applications identified the need to consider communication processes explicitly and equally to the computing tasks. Following this observation, we propose a new communication-aware model for cloud computing applications, called CA-DAG. This model is based on Directed Acyclic Graphs (DAGs) that in addition to computing vertices include separate vertices to represent communications. Such a representation allows making separate resource allocation decisions, assigning processors to handle computing jobs and network resources for information transmissions, such as application database requests.

This paper addresses nonclairvoyant and non-preemptive online job scheduling in Grids. In the applied basic model, the grid system consists of a large number of identical processors that are divided into several machines. Jobs are independent, they have a fixed degree of parallelism, and they are submitted over time. Further, a job can only be executed on the processors belonging to the same machine. It is our goal to minimize the total makespan. We show that the performance of Garey and Graham's list scheduling algorithm is significantly worse in grids than in multiprocessors. Then we present a Grid scheduling algorithm that guarantees a competitive factor of 5. This algorithm can be implemented using a "job stealing" approach and may be well suited to serve as a starting point for Grid scheduling algorithms in real systems.

The problem of multiple sequence alignment is one of the most important problems in computational biology. In this paper we present a new method that simultaneously performs multiple sequence alignment and phylogenetic tree inference for large input data sets. We describe a parallel implementation of our method that utilises simulated annealing metaheuristic to find locally optimal phylogenetic trees in reasonable time. To validate the method, we perform a set of experiments with synthetic as well as real-life data