A privacy preserving framework for federated learning in smart healthcare systems

Federated Learning (FL) is a platform for smart healthcare systems that use wearables and other Internet of Things enabled devices. However, source inference attacks (SIAs) can infer the connection between physiological data in training datasets with FL clients and reveal the identities of participants to the attackers. We propose a comprehensive smart healthcare framework for sharing physiological data, named FRESH, that is based on FL and ring signature defense from the attacks. In FRESH, physiological data are collected from individuals by wearable devices. These data are processed by edge computing devices (e.g., mobile phones, tablet PCs) that train ML models using local data. The model parameters are uploaded by edge computing devices to the central server for joint training of FL models of disease prediction. In this procedure, certificateless ring signature is used to hide the source of parameter updates during joint training for FL to effectively resist SIAs. In the proposed ring signature schema, an improved batch verification algorithm is designed to leverage additivity of linear operations on elliptic curves and to help reduce the computing workload of the server. Experimental results demonstrate that FRESH effectively reduces the success rate of SIAs and the batch verification method significantly improves the efficiency of signature verification. FRESH can be applied to large scale smart healthcare systems with FL involving large numbers of users.     

A Big Data system supporting Bosch Braga Industry 4.0 strategy

People, devices, infrastructures and sensors can constantly communicate exchanging data and generating new data that trace many of these exchanges. This leads to vast volumes of data collected at ever increasing velocities and of different variety, a phenomenon currently known as Big Data. In particular, recent developments in Information and Communications Technologies are pushing the fourth industrial revolution, Industry 4.0, being data generated by several sources like machine controllers, sensors, manufacturing systems, among others. Joining volume, variety and velocity of data, with Industry 4.0, makes the opportunity to enhance sustainable innovation in the Factories of the Future. In this, the collection, integration, storage, processing and analysis of data is a key challenge, being Big Data systems needed to link all the entities and data needs of the factory. Thereby, this paper addresses this key challenge, proposing and implementing a Big Data Analytics architecture, using a multinational organisation (Bosch Car Multimedia – Braga) as a case study. In this work, all the data lifecycle, from collection to analysis, is handled, taking into consideration the different data processing speeds that can exist in the real environment of a factory (batch or stream).     

Modeling and validation of autoinducer-mediated bacterial gene expression in microfluidic environments

Biosensors exploiting communication within genetically engineered bacteria are becoming increasingly important for monitoring environmental changes. Currently, there are a variety of mathematical models for understanding and predicting how genetically engineered bacteria respond to molecular stimuli in these environments, but as sensors have miniaturized towards microfluidics and are subjected to complex time-varying inputs, the shortcomings of these models have become apparent. The effects of microfluidic environments such as low oxygen concentration, increased biofilm encapsulation, diffusion limited molecular distribution, and higher population densities strongly affect rate constants for gene expression not accounted for in previous models. We report a mathematical model that accurately predicts the biological response of the autoinducer N-acyl homoserine lactone-mediated green fluorescent protein expression in reporter bacteria in microfluidic environments by accommodating these rate constants. This generalized mass action model considers a chain of biomolecular events from input autoinducer chemical to fluorescent protein expression through a series of six chemical species. We have validated this model against experimental data from our own apparatus as well as prior published experimental results. Results indicate accurate prediction of dynamics (e.g., 14% peak time error from a pulse input) and with reduced mean-squared error with pulse or step inputs for a range of concentrations (10 μM–30 μM). This model can help advance the design of genetically engineered bacteria sensors and molecular communication devices.     

On the feasibility of Federated Learning towards on-demand client deployment at the edge

Nowadays, researchers are investing their time and devoting their efforts in developing and motivating the 6G vision and resources that are not available in 5G. Edge computing and autonomous vehicular driving applications are more enhanced under the 6G services that are provided to successfully operate tasks. The huge volume of data resulting from such applications can be a plus in the AI and Machine Learning (ML) world. Traditional ML models are used to train their models on centralized data sets. Lately, data privacy becomes a real aspect to take into consideration while collecting data. For that, Federated Learning (FL) plays nowadays a great role in addressing privacy and technology together by maintaining the ability to learn over decentralized data sets. The training is limited to the user devices only while sharing the locally computed parameter with the server that aggregates those updated weights to optimize a global model. This scenario is repeated multiple rounds for better results and convergence. Most of the literature proposed client selection methods to converge faster and increase accuracy. However, none of them has targeted the ability to deploy and select clients in real-time wherever and whenever needed. In fact, some mobile and vehicular devices are not available to serve as clients in the FL due to the highly dynamic environments and/or do not have the capabilities to accomplish this task. In this paper, we address the aforementioned limitations by introducing an on-demand client deployment in FL offering more volume and heterogeneity of data in the learning process. We make use of containerization technology such as Docker to build efficient environments using any type of client devices serving as volunteering devices, and Kubernetes utility called Kubeadm to monitor the devices. The performed experiments illustrate the relevance of the proposed approach and the efficiency of the deployment of clients whenever and wherever needed.     

Multimodal data as a means to understand the learning experience

Most work in the design of learning technology uses click-streams as their primary data source for modelling & predicting learning behaviour. In this paper we set out to quantify what, if any, advantages do physiological sensing techniques provide for the design of learning technologies. We conducted a lab study with 251 game sessions and 17 users focusing on skill development (i.e., user's ability to master complex tasks). We collected click-stream data, as well as eye-tracking, electroencephalography (EEG), video, and wristband data during the experiment. Our analysis shows that traditional click-stream models achieve 39% error rate in predicting learning performance (and 18% when we perform feature selection), while for fused multimodal the error drops up to 6%. Our work highlights the limitations of standalone click-stream models, and quantifies the expected benefits of using a variety of multimodal data coming from physiological sensing. Our findings help shape the future of learning technology research by pointing out the substantial benefits of physiological sensing.     

一种土壤湿度检测系统设计

本文设计了一种无线土壤湿度采集装置,该装置通过湿度传感器HS1101对土壤湿度进行采集,然后将采集到的数据通过无线模块PTR2000发送出去,接收方接收到数据后解包,计算出湿度值并在液晶屏上显示。系统发射部分以AT89S52为核心,包括湿度采集、无线发射。接收部分也以AT89S52为核心,将无线接收、液晶显示结合起来,通过适当的软、硬件抗干扰处理,该系统具有实用性、小型化等特点。     

Mining social influence in science and vice-versa: A topic correlation approach

There is no doubt that scientific discoveries have always brought changes to society. New technologies help solve social problems such as transportation and education, while research brings benefits such as curing diseases and improving food production. Despite the impacts caused by science and society on each other, this relationship is rarely studied and they are often seen as different universes. Previous literature focuses only on a single domain, detecting social demands or research fronts for example, without ever crossing the results for new insights. In this work, we create a system that is able to assess the relationship between social and scholar data using the topics discussed in social networks and research topics. We use the articles as science sensors and humans as social sensors via social networks. Topic modeling algorithms are used to extract and label social subjects and research themes and then topic correlation metrics are used to create links between them if they have a significant relationship. The proposed system is based on topic modeling, labeling and correlation from heterogeneous sources, so it can be used in a variety of scenarios. We make an evaluation of the approach using a large-scale Twitter corpus combined with a PubMed article corpus. In both of them, we work with data of the Zika epidemic in the world, as this scenario provides topics and discussions on both domains. Our work was capable of discovering links between various topics of different domains, which suggests that some of the relationships can be automatically inferred by the sensors. Results can open new opportunities for forecasting social behavior, assess community interest in a scientific subject or directing research to the population welfare.     

航空数据转发设备仿真器研究与设计

航空数据转发设备仿真器是用于飞机上的多台仿真设备（如惯导卫星传感器、气象雷达仿真器等）之间的通信。其中AFDX和ARINC429是新一代航空总线协议,硬件方面需要采用标准的ARINC429通讯卡、API-FDX -2通讯卡。数据转发设备仿真器需要完成实验室对硬线的自由转接和数据仿真工作,可按照系统要求进行节点自由转接以及与其它设备进行AFDX、ARINC429数据交互,以满足某航空研究所实验室的仿真需求。     

The role of big data in smart city

The expansion of big data and the evolution of Internet of Things (IoT) technologies have played an important role in the feasibility of smart city initiatives. Big data offer the potential for cities to obtain valuable insights from a large amount of data collected through various sources, and the IoT allows the integration of sensors, radio-frequency identification, and Bluetooth in the real-world environment using highly networked services. The combination of the IoT and big data is an unexplored research area that has brought new and interesting challenges for achieving the goal of future smart cities. These new challenges focus primarily on problems related to business and technology that enable cities to actualize the vision, principles, and requirements of the applications of smart cities by realizing the main smart environment characteristics. In this paper, we describe the state-of-the-art communication technologies and smart-based applications used within the context of smart cities. The visions of big data analytics to support smart cities are discussed by focusing on how big data can fundamentally change urban populations at different levels. Moreover, a future business model of big data for smart cities is proposed, and the business and technological research challenges are identified. This study can serve as a benchmark for researchers and industries for the future progress and development of smart cities in the context of big data.     

A neural network-based detection and mitigation system for unintended acceleration

Modern vehicles are equipped with a growing number of electronic devices, which significantly improve the driving experience. However, the complicated architecture of electronic systems also increases the difficulty of fault diagnosis since process models are often unavailable. This paper presents a novel detection and mitigation system for vehicle related anomalies originating in unintended acceleration (UA), which has become one of the most complained-about vehicle problems in recent history. The detection system consists of several neural network-based models, which are created by analyzing historical vehicle data at specific moments such as acceleration peaks and gear shifting. These data-driven models describe the boundary of normal vehicle behavior in the data space. A priori knowledge of complete vehicle structures is not necessary for building them. The detection system combines these models to decide if a UA event has occurred. When a UA event is detected, a mitigation system cuts the engine power and adjusts the braking force accordingly. The whole system was validated in the Simulink/dSPACE environment. UA errors were simulated so that they occurred randomly when human subjects drove virtual cars in a simulated environment. Random noise of sensors were also considered and incorporated to add realism. Various traffic scenarios were included in tests. Test results show that the integrated system is capable of detecting UA in one second with high accuracy and reducing the risk of accidents.     

Stretchable and self-healable hydrogel artificial skin

Hydrogels have emerged as promising materials for the construction of skin-like mechanical sensors. The common design of hydrogel-based artificial skin requires a dielectric sandwiched between two hydrogel layers for capacitive sensing. However, such a planar configuration limits the sensitivity, stretchability and self-healing properties. Here, we report the design of single-layer composite hydrogels with bulk capacitive junctions as mechanical sensors. We engineer dielectric peptide-coated graphene (PCG) to serve as homogenously dispersed electric double layers in hydrogels. Any mechanical motions that alter the microscopic distributions of PCG in the hydrogels can significantly change the overall capacitance. We use peptide self-assembly to render strong yet dynamic interfacial interactions between the hydrogel network and graphene. The resulting hydrogels can be stretched up to 77 times their original length and self-heal in a few minutes. The devices can effectively sense strain and pressure in both air and aqueous environments, providing tremendous opportunities for next-generation iontronics.     

Anisotropic fluoride nanocrystals modulated by facet-specific passivation and their disordered surfaces

Rutile-type fluorides have been proven to be active components in the context of emerging antiferr-omagnetic devices. However, controlled synthesis of low-dimensional, in particular two-dimensional (2D), fluorides in a predictable and deterministic manner remains unrealized because of a lack of efficient anisotropic control, which impedes their further development in reduced dimensions. We report here that altered passivation of {110} growing facets can direct the synthesis of rutile-type fluoride nanocrystals into well-defined zero-dimensional (0D) particulates, one-dimensional (1D) rods and 2D sheets in a colloidal approach. The obtained nanocrystals show positive exchange bias and enhanced magnetic transition temperature from the coexistence of long-range antiferromagnetic order and disordered surface spins, making them strong alternatives for flexible magnetic devices and sensors.     

Complex adaptive filtering user profile using graphical models

This article explores how to develop complex data driven user models that go beyond the bag of words model and topical relevance. We propose to learn from rich user specific information and to satisfy complex user criteria under the graphical modelling framework. We carried out a user study with a web based personal news filtering system, and collected extensive user information, including explicit user feedback, implicit user feedback and some contextual information. Experimental results on the data set collected demonstrate that the graphical modelling approach helps us to better understand the complex domain. The results also show that the complex data driven user modelling approach can improve the adaptive information filtering performance. We also discuss some practical issues while learning complex user models, including how to handle data noise and the missing data problem.     

An integrated model for textual social media data with spatio-temporal dimensions

GPS-enabled devices and social media popularity have created an unprecedented opportunity for researchers to collect, explore, and analyze text data with fine-grained spatial and temporal metadata. In this sense, text, time and space are different domains with their own representation scales and methods. This poses a challenge on how to detect relevant patterns that may only arise from the combination of text with spatio-temporal elements. In particular, spatio-temporal textual data representation has relied on feature embedding techniques. This can limit a model’s expressiveness for representing certain patterns extracted from the sequence structure of textual data. To deal with the aforementioned problems, we propose an Acceptor recurrent neural network model that jointly models spatio-temporal textual data. Our goal is to focus on representing the mutual influence and relationships that can exist between written language and the time-and-place where it was produced. We represent space, time, and text as tuples, and use pairs of elements to predict a third one. This results in three predictive tasks that are trained simultaneously. We conduct experiments on two social media datasets and on a crime dataset; we use Mean Reciprocal Rank as evaluation metric. Our experiments show that our model outperforms state-of-the-art methods ranging from a 5.5% to a 24.7% improvement for location and time prediction.     

可扩展的数据采集软件系统开发与设计

当今社会,物联网迅速发展,传感器已在各个领域大量使用,智慧地球的概念正逐步影响个人、企业乃至政府的行为工作方式.而数据采集技术作为物网络一个关键环节,影响着物联网的效率和精确性.在当前的生产和生活当中,一个物联网的建立常常涉及到很多传感器或者数据采集设备的应用,随着科技的发展这些设备的性能和功能也在快速的更新.一个传感器或者数据采集设备的增减常常又会伴随着数据采集程序的更新,企业要与时俱进、提高效率,当随时改善物联网结构,这样的一种改善必然引起新的工程开发,从而导致了新项目实施的成本增加和周期延长.为了避免这种程序的重复开发,提高物联网改善的效率,本文采用在Myeclipse开发平台运用java语言设计并实现了可配置数据采集软件系统.该系统具有高通用性和高移植性,可以通过不同参数配置对不同的传感器或数据采集设备实现数据的采集.     

A Machine-Learning Based Approach for Predicting Older Adults’ Adherence to Technology-Based Cognitive Training

Adequate adherence is a necessary condition for success with any intervention, including for computerized cognitive training designed to mitigate age-related cognitive decline. Tailored prompting systems offer promise for promoting adherence and facilitating intervention success. However, developing adherence support systems capable of just-in-time adaptive reminders requires understanding the factors that predict adherence, particularly an imminent adherence lapse. In this study we built machine learning models to predict participants’ adherence at different levels (overall and weekly) using data collected from a previous cognitive training intervention. We then built machine learning models to predict adherence using a variety of baseline measures (demographic, attitudinal, and cognitive ability variables), as well as deep learning models to predict the next week's adherence using variables derived from training interactions in the previous week. Logistic regression models with selected baseline variables were able to predict overall adherence with moderate accuracy (AUROC: 0.71), while some recurrent neural network models were able to predict weekly adherence with high accuracy (AUROC: 0.84-0.86) based on daily interactions. Analysis of the post hoc explanation of machine learning models revealed that general self-efficacy, objective memory measures, and technology self-efficacy were most predictive of participants’ overall adherence, while time of training, sessions played, and game outcomes were predictive of the next week's adherence. Machine-learning based approaches revealed that both individual difference characteristics and previous intervention interactions provide useful information for predicting adherence, and these insights can provide initial clues as to who to target with adherence support strategies and when to provide support. This information will inform the development of a technology-based, just-in-time adherence support systems.     

Machine learning-enabled multiplexed microfluidic sensors

High-throughput, cost-effective, and portable devices can enhance the performance of point-of-care tests. Such devices are able to acquire images from samples at a high rate in combination with microfluidic chips in point-of-care applications. However, interpreting and analyzing the large amount of acquired data is not only a labor-intensive and time-consuming process, but also prone to the bias of the user and low accuracy. Integrating machine learning (ML) with the image acquisition capability of smartphones as well as increasing computing power could address the need for high-throughput, accurate, and automatized detection, data processing, and quantification of results. Here, ML-supported diagnostic technologies are presented. These technologies include quantification of colorimetric tests, classification of biological samples (cells and sperms), soft sensors, assay type detection, and recognition of the fluid properties. Challenges regarding the implementation of ML methods, including the required number of data points, image acquisition prerequisites, and execution of data-limited experiments are also discussed.     

On exploiting propagation delays for passive target localization using bearings-only measurements

For many cases, target localization accuracy can significantly improve by accounting for the propagation delay between the target source and the sensors. This paper considers when it is advantageous to compensate for the propagation delay for the case of a network of nodes where each node consists of an array of sensors. Each node calculates a direction of arrival (DOA) from the raw data, and these DOAs are combined to localize the target. First, we consider the case when no prior data are given and the localization occurs using DOA measurements from a single snapshot. Such localization methods are necessary to initiate a target track. Finally, we investigate the case where an extended Kalman filter is used to aggregate measurements over multiple snapshots.     

Paper-based colorimetric enzyme linked immunosorbent assay fabricated by laser induced forward transfer

We report the Laser Induced Forward Transfer (LIFT) of antibodies from a liquid donor film onto paper receivers for application as point-of-care diagnostic sensors. To minimise the loss of functionality of the active biomolecules during transfer, a dynamic release layer was employed to shield the biomaterial from direct exposure to the pulsed laser source. Cellulose paper was chosen as the ideal receiver because of its inherent bio-compatibility, liquid transport properties, wide availability and low cost, all of which make it an efficient and suitable platform for point-of-care diagnostic sensors. Both enzyme-tagged and untagged IgG antibodies were LIFT-printed and their functionality was confirmed via a colorimetric enzyme-linked immunosorbent assay. Localisation of the printed antibodies was exhibited, which can allow the creation of complex 2-d patterns such as QR codes or letters for use in a final working device. Finally, a calibration curve was determined that related the intensity of the colour obtained to the concentration of active antibodies to enable quantitative assessment of the device performance. The motivation for this work was to implement a laser-based procedure for manufacturing low-cost, point-of-care diagnostic devices on paper.     

What IoT devices and applications should be connected? Predicting user behaviors of IoT services with node2vec embedding

User-created automation applets to connect IoT devices and applications have become popular and widely available. Exploring those applets enables us to grasp the patterns of how users are utilizing and maximizing the power of connection by themselves, which can deliver practical implications for IoT service design. This study builds an IoT application network with the data of the IFTTT(if this then that) platform which is the most popular platform for self-automation of IoT services. The trigger-action relationships of the IFTTT applets currently activated are collected and used to construct an IoT application network whose nodes are IoT service channels, and links represent their connections. The constructed IoT network is then embedded by the node2vec technique, an algorithmic framework for representational learning of nodes in networks. Clustering the embedded nodes produces the four clusters of IoT usage patterns: Smart Home, Activity Tracking, Information Digest, and Lifelogging & Sharing. We also predict the IoT application network using node2vec-based link prediction with several machine learning classifiers to identify promising connections between IoT applications. Feasible service scenarios are then generated from predicted links between IoT applications. The findings and the proposed approach can offer IoT service providers practical implications for enhancing user experiences and developing new services.