With the expansion of embedded systems and the new applications of these computing systems in Internet of Things, smart homes and sensors, such systems are faced with a large amount of data. Therefore, designing an embedded system for Massive sensor data analysis for detecting a situation and making a decision is becoming more complex. Machine learning algorithms are a promising method for designing such systems. Neural networks are very popular learning systems to be used in data analysis. However, huge computations of the neural networks make them unsuitable for embedded systems which have strict limitations on energy consumption and execution time as the main design concerns. In this paper we propose an optimization for multi-layer perceptron neural networks based on neuron saturation states. Experimental results show that in average 89% of neurons are in saturated state for benchmark applications. Since in a saturated neuron, exact numerical value is not required, predicting the saturation state helps to perform the computation with less effort. The goal of the saturation prediction is to keep accuracy while minimizing the computations.

Smart vehicles are one of the applications of the Internet of Things. Controller area network (CAN) is the standard mean for its communication. The security of CAN is important and may be prone to side channel attacks. DPA is one of such attacks, which is considered among the most serious threats against cryptographic devices. Various metrics have been proposed to evaluate the resistance of different implementations against these attacks. Some of these metrics need side-channel attacks to be conducted and depend on the considered power model. Due to the vast variety of proposed side-channel attacks and power models, comprehensively evaluating a design under these metrics is commonly considered to be a tedious task. To alleviate this situation, t-test has been proposed. The non-specific variation of t-test does not need to suppose a power model for leakage assessment. In this paper, we have evaluated three implementations of AES on an FPGA. In the first design, no side-channel countermeasure is implemented while the second and the third implementations make use of masking and shuffling, respectively. Evaluation results show that significant reduction occurs in terms of side-channel leakage when masking or shuffling is applied. Results imply that shuffling and masking are proper choices for area-restricted and time-restricted devices, respectively.

With the widespread use of Internet of Things (IoT) in every aspect of human’s daily life, communications of such an enormous amount of existing embedded devices in these systems arise many new challenges from power consumption, performance, and reliability perspectives. Communications in an IoT infrastructure are managed by a set of policies which are determined by Objective Functions (OFs). Thus, OFs are the most important contributors in facing with the mentioned challenges. In this paper, due to the lack of information on how OFs affect the primary properties of an IoT infrastructure, we have compared three well-known OFs (OF0, MRHOF, and OFFL) from power consumption, performance, and reliability perspectives. This paper would pave the way for scholars to realize which of the three mentioned areas are more affected by OFs. Our evaluation results on Contiki’s IoT simulation environment (COOJA) illustrate, OFs could affect the power consumption by up to 71.2%, while their impacts on reliability and performance would be as high as 77.5% and 25.15%, respectively.

Software defined networking and network virtualization are widely considered promising techniques for reducing the complexity of network management in many contexts that require high Quality of Service (QoS) and the support for heterogeneous architectures. In this paper we address a network architecture, here called a virtualized SDN network that combines the benefits of SDN and virtualization. To cope with the demand for efficiently sharing a platform among several services, here a resource management mechanism to reserve and control network resources among various services in the virtualized SDN networks is proposed. The mechanism is implemented on an SDN controller and a set of experiments show the effectiveness of the proposed approach.

This paper concentrates on cloud computing as one of the principal building blocks of a smart factory to provide a huge data storage space and a highly scalable computational capacity. The cloud computing system used in a smart factory should be a time-predictable computing framework to be able to satisfy hard-real requirements of various applications existing in manufacturing systems. Interleaving an intermediate computing layer –called Fog– between a factory and a cloud data center is a promising solution for such a system to deal with the latency requirements of hard real-time applications. In this paper, a time-predictable cloud framework is proposed which is able to satisfy end-to-end latency requirements in a smart factory. To propose such an industrial cloud framework, we not only use the existing real-time technologies such as Industrial Ethernet and Real-time XEN hypervisor, but we also discuss unaddressed challenges. Among the unaddressed challenges, the partitioning of a given workload between a fog platform and a cloud data center is targeted. Addressing the partitioning problem not only provides a resource provisioning mechanism, but also gives us a prominent design decision to know how much computing resource is required to develop a Fog platform, and how large should be the minimum communication bandwidth between a fog and a cloud data center.

Development of Internet of things (IoT) applications brings a new movement to the functionality of wireless sensor networks (WSNs) from only environment sensing and data gathering to collaborative inferring and ubiquitous intelligence. In intelligent WSNs, nodes collaborate to exchange the information needed to achieve the required inference or smartness. Efficiency or correctness of many smart applications rely on efficient, timely, reliable, and ubiquitous inference of information which may continually change in interaction with other nodes. In this paper, we introduce a framework which provides a generic solution for such common ubiquitous inferences. this framework brings the reliability and ubiquity for inferences using the redundancy characteristic of the gossiping protocols. With this framework the implementation of such inferences and the control of their speed and cost is abstracted by providing the developers with a proposed middleware and some dissemination control services. We develop a prototype implementation of the proposed framework and a few inference examples on TinyOS to evaluate the effectiveness of the framework and the efficiency of the inferences. To this aim, we utilize TOSSIM simulator in various densities and different number of nodes. Results of the evaluations demonstrate that in all nodes, the inferring time after a change is about a few seconds (less than a minute) and the cost of maintenance in steady state is about a few sends (less than 10) per hour.

The emission of radio waves from Extensive Air Showers (EAS), initiated by ultrahigh-energy cosmic rays, has been attributed to geomagnetic emission and charge excess processes. At frequencies 10-100 MHz this process leads to coherent radiation. Nowadays, the radio detection technique is used in many experiments consisting in studying EAS. One of them is the Auger Engineering Radio Array (AERA), located within the Pierre Auger Observatory. The frequency band observed by AERA radio stations is 30-80 MHz. The investigated frequency range is often very contaminated by human-made and narrowband radio frequency interferences (RFI). Suppression of these contaminations is crucial to lower the rate of spurious triggers. An adaptive filter based on the Least Mean Squares (LMS) algorithm can be an alternative one for currently used IIR-notch non-adaptive filter. The paper presents the 32/64-stage filters based on non-canonical FIR filter implemented into cost-effective Cyclone IV and Cyclone V Altera FPGAs with a sufficient safety margin of the registered performance for the global clock above 200 MHz to obey the Nyguist requirement.

Usually real-time control systems are designed by periodic con-trollers with conservative period selection to guarantee the plant stability and control performance. However, in resource-constrained systems, the performance may temporarily degrade due to disturbances, a matter which could be better managed if there were sufficient resources to select smaller control periods. In this paper, we present a dynamic resource management ap-proach to efficiently distribute the processing power among con-current control tasks to improve the overall system control per-formance. The key idea is to use the self-triggered control ap-proach to postpone the trigger of control tasks for non-disturbed plants and allocate the freed resources to run addi-tional instances of the control tasks of the disturbed plants. We show that this idea improves the overall control performance of the disturbed plants in different scenarios.

The heterogeneity in cyber-physical systems (CPS) and the diverse situations that they may face with, along with the environmental hazards raise the need to self-stabilization. The uncertain nature of CPS necessitates a probabilistic view for analyzing the system stabilization-time that is a highly critical metric in distributed/time-sensitive applications. Calculating the worst-case expected stabilization-time and possible improvements help to have safer designs of CPS applications. In this paper, a mutual exclusion algorithm based on PIF (Propagation of Information with Feedback) self-stabilizing algorithm is selected in synchronous environment as a case study. Using probabilistic analysis, we present a set of guidelines for utilizing this algorithm in time-sensitive applications. We have also utilized an approximation method for improving the scalability of our probabilistic analysis and did a set of experiments to show how this analysis could be used in the design of topologies with the goal of having an optimal worst-case expected stabilization-time. Our results show that using this approach, we can significantly improve the worst-case expected stabilization-time.

In this paper, the development of a real-time hardware-in-the-loop (HIL) test for a spacecraft is investigated. The spacecraft mission is to transfer a satellite from 400km circular to 400-7000km elliptical orbit by a small upper stage. The main parts of the mission are de-tumbling, pointing, spin-up, and orbital transfer maneuvers. The developed HIL test aims at demonstrating the performance of the guidance, navigation, and control (GNC) algorithms, simulating the mission operation plan (MOP), completing the end-to-end hardware verification, and testing throughout the whole mission. In the presented framework, four modules are considered: 1) Onboard computer (OBC) and onboard hardware (OBH) module; 2) Onboard software (OBS) module; 3) Flight simulator (FS) module; 4) Logger and command computer (LCC) module. The GNC algorithms and MOP are programmed in the OBC while it communicates with a six degrees of freedom flight simulator, implemented in C# language in the visual studio environment, via CAN bus. The HIL test is performed while sensors have been substituted with dummy thermal samples to simulate their energy consumption function, because the environmental conditions of space and the sensors dependence on the current attitude cannot be duplicated in the laboratory environment. The FS module can be executed in a real-time approach with the frequency of 1 kHz and even faster. All results have been proven to be satisfactory with respect to orbit transfer mission and consequently, the onboard hardware and software are verified to be reliable enough after several repeats of the test.

Despite great advances in different software testing areas, one important challenge, achieving an automated test oracle, has been overlooked by academia and industry. Among various approaches for constructing a test oracle, machine learning techniques have been successful in recent years. However, there are some situations in which the existing machine learning based oracles have deficiencies. These situations include testing of applications with low observability, such as embedded software and multimedia software programs. There are also cases in testing embedded software in which explicit historical data in form of input-output relationships is not available, and situations in which the comparison between expected results and actual outputs is impossible or hard. Addressing these deficiencies, this paper proposes a new black box solution to construct automated oracles which can be applied to embedded software and other programs with low observability. To achieve this, we employ an Artificial Neural Network (ANN) algorithm to build a model which only requires input values, and the program’s corresponding pass/fail behaviors, as the training set. We have conducted extensive experiments on several benchmarks. The results manifest a higher accuracy of the proposed approach comparing with a well-known machine learning based method.

Real-time systems are increasingly coming to be implemented in multiprocessor and multicore platforms. In order to achieve full performance gain on these platforms, there is a need for an efficient and scalable implementation. One possible source of inefficiency in these platforms is the shared data structure used for interaction and coordination between threads. In order to prevent race condition resulting from concurrent access to these shared data structures, a locking mechanism is usually used, which while providing safety, limits the performance gain, as at any time, data structure can be accessed by at most one thread of execution. Concurrent data structures try to address these issues. In this work shared data structure used in the context of a real-time multicore scheduling in a real-time operating system is changed to a concurrent version to achieve improved performance and scalability in these platforms.

In this paper, we present an online method in unidirectional IoT systems to synchronize the data streams of multiple sensors, including wearable and environmental. Our proposed method is based on information theory concepts. First we use an entropy-based method to find events on environmental sensors data stream. After that, we use mutual information matching algorithm to find corresponding event in wearable sensors data stream. Indeed, we use physical interactions between wearable and environmental sensors in order to find shared events. Our experiments demonstrate that our proposed method can improve the quality of sensors data stream and therefore it can be used for the monitoring task in unidirectional IoT systems.

Recent trends in real-time systems are towards multicores and parallel processes in the form of directed acyclic graphs. The scheduling aspect of such systems has been worked on and many methods are developed. Nevertheless, the need for more efficient approaches which can use fewer number of cores has not vanished. Semi-partitioned scheduling of hard real-time parallel tasks in multicores is studied in this paper. Since there is no benefit in completing a task much before its deadline, after scheduling a parallel task, if it is beneficial for other tasks, the execution of this task is further moved towards its deadline, i.e., stretching, to make room for tasks with closer deadlines. A new concept, prior+, load of tasks is used to rank all tasks of each directed acyclic graph and order them for scheduling. The scheduler is offline and the schedule map is used during run time. One benefit of this is the reduction of the scheduling overhead during run time which helps to safely accept loads. The comparative evaluations show the algorithms performance is superior to the state of the art ones. It also confirmed that the new concept of prior+ load of a task is very effective in scheduling real-time directed acyclic graphs and suggests that it can as well be useful in scheduling workflows.

In recent years, Android have become a popular operating system for smartphones. Regarding this growing trend, a wide variety of applications are being developed for this platform. Due to limited energy supply on such smartphones, the energy consumption of an application plays a significant role in users’ satisfaction. Recent studies in this field reveals that a considerable number of applications suffer from energy issues. To this end, we are going to provide a systematic framework for application developers aiming to improve the level of energy consumption in application. In this research, we elaborate a profiling technique in order to model users’ behavior in accordance with how they use the application.

يکي از مسائل مهم در شبکه‌هاي اجتماعي، تشخيص جوامع است که در سال‌هاي اخير روش‌هاي متعددي براي آن ارائه شده است. نظريه بازي‌ها از روش‌هاي نوين در اين زمينه است. در اين مقاله، مسئله تشخيص جوامع هم‌پوشان با استفاده از بازي تشکيل ائتلاف مدل و سپس حل شده است. در جوامع هم‌پوشان هر فرد مي‌تواند عضو چندين گروه باشد. در اين راستا الگوريتمي سلسله مراتبي مبتني بر بازي تشکيل ائتلاف ارائه شده است. از عمده‌ترين نقاط ضعف الگوريتم مورد استفاده، وابستگي بيش از حد آن به يکي از پارامترها مي‌باشد به نحوي که براي ديتاست‌هاي مختلف مقادير متفاوتي دارد. در اين مقاله با بکارگيري معيار فاصله وابستگي به اين پارامتر از بين رفته است. معيار فاصله در تمامي فرمول‌ها تاثير داده شده است و براي اولين بار بازي تشکيل ائتلاف در تشخيص جوامع با استفاده از مفهوم فاصله توسعه داده شده است. علاوه بر از بين بردن حساسيت به پارامتر در الگوريتم جديد، آزمايش‌هاي انجام شده نشان مي‌دهد که کيفيت جوامع تشخيص داده شده به طور ميانگين ۴ درصد بهتر شده است.

تشخيص جامعه در شبکه‌هاي اجتماعي از نظر محاسباتي کاري چالش برانگيز است و توجه محققان زيادي را در دهه اخير به خود جلب کرده است. بيشتر مقالات در اين حوزه بر مدل‌سازي ويژگي‌هاي ساختاري تمرکز کرده‌اند. در اکثر اين مقالات، روش‌ها جوامع را يکي پس از ديگري و به صورت سري يا نوبتي تشخيص مي‌دهند. با هدف افزايش سرعت محاسبات، الگوريتم‌هاي تشخيص جامعه موازي‌سازي شدند. الگوريتم انتشار برچسب (LPA) به دليل بازدهي زماني، روش مؤثري براي موازي‌سازي است. در اين مقاله، اين الگوريتم به صورتي بازنويسي شده است که جوامع همپوشان را تشخيص دهد؛ در جوامع همپوشان يک فرد/گره مي‌تواند همزمان به بيش از يک جامعه متعلق باشد. همچنين، مبتني بر مدل نگاشت - کاهش الگوريتم جديد پياده‌سازي شده است. آزمايش‌هاي طراحي شده نشان مي‌دهد که الگوريتم پياده‌سازي شده از الگوريتم‌هاي مشابه تشخيص جوامع همپوشان مانند LPA، SLPA و COPRA براساس معيار ماژولاريتي کارکرد بهتري دارد.

گزارش‌هاي خبري در رسانه‌هاي اجتماعي و خبري به وسيله تعداد زيادي سند ارائه مي‌شوند و شامل موضوعاتي هستند که جوامع و نظرات مختلف را در برمي‌گيرند. آگاهي از رابطه‌‌ ميان افراد در اسناد مي‌تواند به خوانندگان کمک کند تا يک دانش اوليه در‌خصوص موضوع و هدف در اسناد مختلف به‌دست آورند. در پژوهش حاضر، به بررسي مسئله‌ي تشخيص جوامعي پرداخته خواهد شد که هدف اصلي، خوشه‌بندي افرادي است که نام آن‌ها در مجموعه‌اي از اسناد خبري آورده شده است؛ اين افراد در جوامعي با مواضع مرتبط، خوشه‌بندي مي‌شوند. در اين پژوهش يک روش تشخيص جوامع موضع افراد بهبود يافته مبتني بر شبکه دوستي به نام SCIFNET ارائه شده است. روش پيشنهادي از ساختار الگوريتم ژنتيک جهت اين مهم استفاده نموده است. در آزمايش‌ها معيار صحت به منظور مقايسه استفاده شده است که براي رسيدن به اين مهم از شاخص Rand index (که يک شاخص مهم در ارزيابي کيفيت خوشه‌بندي است) استفاده شده است. نتايج حاصل از آزمايش‌ها که برمبناي پايگاه‌هاي داده‌ي واقعي (اسناد منتشر يافته در رسانه خبري گوگل نيوز در رابطه با يک موضوع خاص) به‌ دست‌ آمده‌اند، حاکي از کارآمدي و بهره‌وري مطلوب روش پيشنهادي است.

بازيابي تصاوير يکي از مهم‌ترين مسائل مطرح شده در حوزه بينايي ماشين مي‌باشد که کاربرد گسترده‌اي در علوم مختلف دارد. چالش‌هاي فراواني در زمينه بازيابي تصاوير براساس محتوا وجود دارد، از جمله آنها مي‌توان به وجود نويز، پس زمينه پيچيده و عدم وجود نور کافي در تصاوير اشاره کرد. براي غلبه بر اين مشکلات بايد ويژگي‌هاي استخراج شده قابليت تمايز خوبي داشته باشند. در اين مقاله از الگوريتم کدگذاري خطي با قيد محلي به منظور استخراج ويژگي از تصاوير براي بازيابي استفاده شده است. در ابتدا ويژگي‌هاي SIFT از تصاوير پايگاه داده استخراج مي‌شوند. سپس ويژگي‌هاي استخراج شده خوشه‌بندي شده و مراکز خوشه‌ها به عنوان نماينده هر دسته انتخاب مي‌شود. پس از خوشه‌بندي توصيفگرها، از ضرايب بازنمايي توصيفگرهاي هر دسته با توجه به نماينده‌هاي توليد شده از مرحله قبل، از الگوريتم کدگذاري خطي با قيد محلي به عنوان ويژگي استفاده مي‌شود. در نهايت از اين ويژگي‌هاي توليد شده براي بازيابي تصاوير استفاده خواهد شد. نتايج آزمايشات روي پايگاه داده Caltech-101 برتري روش ارائه شده را در بازيابي تصاوير نسبت به روش‌هاي مطرح در اين حوزه نشان مي‌دهد.

با گسترش روزافزون اينترنت اشياء نقش مهم پروتکل مسيريابي شبکه‌هاي کم‌توان و پُر اتلاف، که توسط کارگروه مهندسي اينترنت به‌عنوان يک استاندارد مسيريابي ويژه معرفي گرديده، بيشتر شده است. به‌جهت محدوديت‌هاي ذاتي شبکه‌هاي مبتني بر RPL، اين شبکه‌ها در معرض انواع حملات قرار مي‌گيرند. حتي اگر از مکانيزم‌هاي رمزنگاري به‌عنوان اولين خط دفاعي استفاده شود تنها مي‌توان از حملات خارجي محافظت نمود. هنگامي که گره‌ها به‌عنوان يک مهاجم داخلي در نظر گرفته شوند تکنيک‌هاي رمزنگاري نمي‌توانند هيچ‌گونه محافظتي از شبکه‌ها انجام دهند. يکي از مهم‌ترين و جدي‌ترين اين نوع حملات، حمله سياه‌چاله است که در آن يک گره مهاجم همه بسته‌هايي که تصور مي‌گردد ارسال خواهند شد را حذف مي‌نمايد. معيار مصرف انرژي به‌عنوان يکي از مهم‌ترين محدوديت‌هاي شبکه‌هاي RPL در روش‌هاي ارائه شده کنوني براي مقابله با اين حمله در نظر گرفته نشده است. در اين مقاله يک مکانيزم انرژي‌کارا جهت کاهش اثرات اين حمله با بيشترين نرخ دريافت بسته‌هاي داده و دقت بالا ارائه مي‌گردد. در اين مکانيزم هر گره رفتار مشکوک گره‌هاي همسايه را با شنود چگونگي ارسال و دريافت بسته‌ها تشخيص مي‌دهد. ارزيابي‌هاي انجام شده توسط سيستم‌عامل کانتيکي نشان مي‌دهد مکانيزم فوق داراي نرخ دريافت بسته بالا، تأخير کم و مصرف انرژي مناسب مي‌باشد.