Search

In this full research paper we examine questionable collaboration from a student perspective. Collaborating while solving computer lab assignments is often considered an important part when learning computer science, as it allows students to discuss their work, while also practicing working together. However, it also introduces risks, such as students collaborating in ways negatively impacting their learning outcomes and leading to inaccurate grading. Hence it is important to work towards reducing the use of these poor collaborative practices. In order to ameliorate the problem with academic misconduct, we need to understand students’ justifications for deviating from acceptable practices. In this paper we therefore investigate how students justify their collaborative practices during computer lab assignments in situations they experience as questionable. The justifications were collected through 15 semi-structured interviews with students experienced in pair programming, majoring in computer science and other technical fields from two large well-known European universities.The justifications from the interviews were analysed using phenomenography resulting in seven categories: external pressure, lack of interest, spending time on the assignment, understanding the end product, contributing to the process, learning from the assignment and reflecting on the purpose of the learning. These describe in which situations students might deviate from the rules and can be used by institutions to prevent such behavior.

Patient monitoring in intensive care units requires collection and processing of high volumes of data. High sensitivity of sensors leads to significant number of false alarms, which cause alarm fatigue. Reduction of false alarms can lead to better reaction time of medical personnel. This paper aims to develop a method for false alarm suppression and evaluate it on a publicly available data set with manually annotated alarms. First, an automated feature engineering was performed using the signal for arterial blood pressure (ABP) and a processed signal that contained the times of each heartbeat from the ABP signal. Next, support vector machines, random forest, and extreme random trees classifiers were trained to create classification models. The best suppression performance was achieved for the extreme tachycardia alarm, for which 90.3% of the false alarms were suppressed, while only 0.54% of the true alarms were incorrectly suppressed. This paper demonstrates that alarm suppression can be achieved with high accuracy using an automated feature engineering coupled with machine learning algorithms. The proposed approach can be utilized as aid to medical personnel and experts, allowing them to be more productive and to respond to alarms in a more timely manner.

Network virtualization enables multi-tenancy over physical network infrastructure, with a side-effect of increased network complexity. Software-defined networking (SDN) is a novel network architectural model – one where the control plane is separated from the data plane by a standardized API – which aims to reduce the network management overhead. However, as the SDN model itself is evolving, its application to multi-tenant virtualized networks raises multiple security challenges. In this paper, we present a security analysis of SDN- based multi-tenant virtualized networks: we outline the security assumptions applicable to such networks, define the relevant adversarial model, identify the main attack vectors for such network infrastructure deployments and finally synthesize a set of high-level security requirements for SDN-based multi-tenant virtualized networks. This paper sets the foundation for future design of secure SDN-based multi-tenant virtualized networks.

The main goal of this paper is to extract the semantic relations underpinning the concepts of English prepositional of-constructions derived from poetic and non-poetic datasets, using Princeton WordNet. The problem is addressed by two different algorithms, which are evaluated for their ability to model the different types of resources from which the relations are derived, and for their ability to predict unseen relations. The first algorithm introduces the concept of subsumption hierarchy between relations in order to derive the most general relations associated to each type of data source and identify a set of relations specific to each dataset. The second algorithm investigates the use of a weighting scheme in order to establish the importance of each association extracted. Of particular importance are the notions of subsumption hierarchies between relations (expressed as synset pairs) and the Inverse Relation Frequency (IRF) measure, which is inspired by the Inverse Document Frequency measure used in Information Retrieval. The ontological prospects of using Princeton WordNet and the above algorithms for the creation of ontologies are also briefly discussed. Although the main interest of the proposed methods lies to the identification of conceptual relations particular to poetic resources, the methods followed can be applied and are evaluated on other domains too.

Internet of Things (IoT) has been found pervasive use cases and become a driving force to constitute a digital society. The ultimate goal of IoT is data and the intelligence generated from data. With the progress in public cloud computing technologies, more and more data can be stored, processed and analyzed in cloud to release the power of IoT. However, due to the heterogeneity of hardware and communication protocols in the IoT world, the interoperability and compatibility among different link layer protocols, sub-systems, and back-end services have become a significant challenge to IoT practices. This challenge cannot be addressed by public cloud suppliers since their efforts are mainly put into software and platform services but can hardly be extended to end devices. In this paper, we propose a data-centric IoT framework that incorporates three promising protocols with fundamental security schemes, i.e., WiFi, Thread, and LoRaWAN, to cater to massive IoT and broadband IoT use cases in local, personal, and wide area networks. By taking advantages of the Azure cloud infrastructure, the framework features a unified device management model and data model to conquer the interoperability challenge. We also provide implementation and a case study to validate the framework for practical applications. Funding agencies: Swedish Environmental Protection Agency; Norrkoping Fund for Research and Development, Sweden

Automatic Image Annotation (AIA) is the process of assigning tags to digital images without the intervention of humans. Most of the modern automatic image annotation methods are based on the learning by example paradigm. In those methods building the training examples, that is, pairs of images and related tags, is the first critical step. We have shown in our previous studies that hashtags accompanying images in social media and especially the Instagram provide a reach source for creating training sets for AIA. However, we concluded that only 20% of the Instagram hashtags describe the actual content of the image they accompany, thus, a series of filtering steps need to apply in order to identify the appropriate hashtags. In this paper we apply graph based topic modelling on Instagram hashtags in order to predict the subject of the related images and we propose an innovative image retrieval scheme that can be used in the context of Instagram with minimal training requirements.

The use of wearable cameras covers several areas of application nowadays, where the need for developing smart applications providing the sustainability and well-being of citizens it is more necessary than ever before. The tremendous amount of lifelogging data to extract valuable knowledge about the every day life of the wearers requires state of the art retrieval techniques to efficiently store, access, search and retrieve useful information. Several works have been proposed combining computer vision and machine learning techniques to analyze the content of the data captured from visual wearable devices on a daily basis. This paper presents an overview of the progress in visual lifelogging retrieval and indicates the current advances and future challenges, highlighting the prospects of incorporating visual lifelogging retrieval in social computing applications.

Randomness beacons are services that periodically emit a random number, allowing users to agree on the same random outcome without trusting anyone: ideally, the randomness beacon is secure (cannot be influenced) and transparent (can be monitored by users). Hence, such randomness beacons can serve as an important primitive for smart contracts in a variety of contexts. In this paper we aim to bridge the gap between theory and practice of public beacon design inspired by the unicorn protocol of Lenstra and Wesolowski using verifiable delay functions. We first present a structured security analysis, based on which we design, implement, and evaluate a trustworthy and efficient randomness beacon allowing users to join at any time. We then compare different implementation and deployment options on distributed ledgers, and report on a Ethereum smart contract-based lottery using our beacon.

Over the last couple of decades, mobile devices have evolved in many ways with powerful processing and high storage capabilities being complemented by a number of sensors and communication interfaces. Exploiting their wireless communication capabilities, mobile devices form intermittently connected networks, known as Mobile Opportunistic Networks. In such networks, devices exchange data packets opportunistically. Dynamic topology changes and difficulty in the routing process often result in long delays with user location data along with social characteristics being potentially useful in the process. In this paper, we propose the deployment of an intelligent network controller, based on the principles of software-defined networks that will collect and process context data from the available wireless networks, users, and social networks. We describe the architecture of our system, analyse component interactions and present the business processes and related application services. We demonstrate the experimental results from a use case scenario for estimating the contact duration between opportunistically connected mobile nodes.

Automatic Image Annotation (AIA) is the process of assigning tags to digital images without the intervention of humans. Most of the modern automatic image annotation methods are based on the learning by example paradigm. In those methods building the training examples, that is, pairs of images and related tags, is the first critical step. We have shown in our previous studies that hashtags accompanying images in social media and especially the Instagram provide a reach source for creating training sets for AIA. However, we concluded that only 20% of the Instagram hashtags describe the actual content of the image they accompany, thus, a series of filtering steps need to apply in order to identify the appropriate hashtags. In this paper we apply topic modelling with Latent Dirichlet Allocation (LDA) on Instagram hashtags in order to predict the subject of the related images. Since a topic is composed by a set of related terms, the identification of the visual topic of an Instagram image, through the proposed method, provides a plausible set of tags to be used in the context of training AIA methods.

The combination of high mobility and wireless communication in many safety-critical systems have increased their exposure to malicious security threats. Consequently, many works in the past have proposed solutions to ensure safety and security of these systems. However, not much attention has been given to the interplay between these two groups of non-functional requirements. This is a concern as safety solutions may negatively impact system security and vice versa. This paper addresses the interplay between safety and security by proposing an attack injection framework, based on model-implemented fault injection, suitable for model-based design. The framework enables us to study and evaluate the impact of cybersecurity attacks on system safety early in the development process. To this end, we have implemented six attack injection models and conducted experiments on Simulink models of a CAN bus and a brake-by-wire controller. The results show that the security attacks modeled could successfully impact the system safety by violating our defined safety requirements.

This paper gives an up-to-date overview of research about Virtual Reality (VR) for inducing empathy and reducing prejudice towards stigmatized groups and the measurements used in the studies. Outcomes from the studies reviewed provide only preliminary support for the use of VR for successfully inducing empathy into people and reducing their prejudice towards stigmatized groups and the preference for using self-report methods for the measurement of empathy and prejudice.

Proactive cognitive agents need to be capable ofboth generating their own goals and enacting them. In thispaper, we cast this problem as that ofmaintaining equilibrium,that is, seeking opportunities to act that keep the system indesirable states while avoiding undesirable ones. We characterizedesirability of states as graded preferences, using mechanismsfrom the field of fuzzy logic. As a result, opportunities for anagent to act can also be graded, and their relative preferencecan be used to infer when and how to act. This paper providesa formal description of our computational framework, andillustrates how the use of degrees of desirability leads to well-informed choices of action., Funding Agencies:Italian Ministry of Foreign Affairs and International Cooperation (MAECI) Italian Ministry of Education, Universities and Research (MIUR) Swedish Research Council (VR Project WearAmI) PGR00193 Semantic Robots Research Profile (Swedish Knowledge Foundation)

Sentiment analysis of Twitter data became a research trend the last decade. Thanks to the Twitter API, massive amounts of tweets, relating to a topic of interest, can be collected in real time. Performing sentiment analysis of these tweets can be used to conduct social sensing and opinion mining. For instance, forecasting elections is a primary area in which sentiment analysis of tweets has been extensively applied the last few years. Sentiment analysis of Twitter data presents important challenges compared to the similar task of text classification. Tweets are limited to 140 characters; thus, the conveyed message is compressed and often context-dependent. The tweets are informal and unstructured, usually lacking grammatical soundness and use of a standard lexicon. On the other hand, tweets are usually annotated by their authors regarding their topic and sentiment with the aid of hashtags and emoticons. Identifying appropriate features for sentiment analysis of tweets remains an open research area since text indexing methods face the sparseness problem while POS tagging methods fail due to the lack of grammatical structure of tweets. Character based features, i.e., n-grams of characters, are currently getting popular because they are language independent. However, their effectiveness remains quite low. In this paper, we argue that tokens used by humans for sentiment analysis of tweets are probably the best feature set one can use for that purpose. We compare several automatically extracted features with the features (tokens) used by humans for tweet classification, under a machine learning framework. The results show that the manually indicated tokens combined with a Decision Tree classifier outperform any other feature set-classification algorithm combination. The manually annotated dataset that was used in our experiments is publicly available for anyone who wishes to use it.

Cloud-based Radio Access Network (C-RAN) is a promising architecture for future cellular networks, in which Baseband Units (BBUs) are placed at a centralized location, with capacity-constrained fronthaul connected to multiple distributed Remote Radio Heads (RRHs) that are far away from the BBUs. The centralization of signal processing enables the flexibility for coordinated multi-point transmission (CoMP) to meet high traffic demand of users. We investigate how to jointly optimize CoMP-cell selection and base station resource allocation so as to enhance the quality of service (QoS), subject to the fronthaul capacity constraint in orthogonal frequency-division multiple access (OFDMA) based C-RAN. The problem is proved to be NP-hard in this paper. To deal with the computational complexity, we derive a partial optimality condition as the foundation for designing a cell-selection algorithm. Besides, we provide a solution method of the optimum of the time-frequency resource allocation problem without loss of fairness on the QoS enhancement of all users. The simulations show good performance of the proposed algorithms for jointly optimizing the cell selection and resource allocation in a C-RAN, with respect to QoS.

The process of learning and understanding can be different for each person. The teaching methods are particularly important for students to learn the curriculum, because not every student can learn the teaching material in the same way. Some students can easily learn if they listen the information, others use videos presentations and graphs for better learning and there are students who learn through practical examples. This paper presents VARK (Visual, Aural, Reading or Write and Kinesthetic) model as an effective learning style. The learning model is applied in the Data structures and algorithms course. Data Structure and Algorithms is a fundamental course in Computer Science, but many students find it difficult because it requires abstract thinking. For this reason, except traditional way of explaining (reading/listening), using visualization tool, watching videos and animations is very helpful. At the University "Goce Delchev" teaching of the subject Data Structures and Algorithms contains: classroom lectures, classroom problem solving exercises and laboratory exercises. Classroom lectures are performed by using only listening and reading methods. On classroom problem solving exercises, students watch videos or animations that are some kind of java applets or Flash elements that illustrate the basic operation in data structure. On laboratory exercises, students use programs for software visualization and have opportunity to write parts of code by their own. This way of combined learning increases overall study experience and improve student success. This method has significant positive effect on students.

Generalized autoregressive conditional heteroscedasticity (GARCH) models have long been considered as one of the most successful families of approaches for volatility modeling in financial return signals. However, this family of methods employ quite rigid assumptions regarding the evolution of the variance. In this paper, we address these issues by introducing a recurrent latent variable model, capable of capturing highly flexible functional relationships for the variances. We derive a fast, scalable, and robust to overfitting Bayesian inference algorithm, by relying on amortized variational inference. This avoids the need to compute per-data point variational parameters, but can instead compute a set of global variational parameters valid for inference at both training and test time. We evaluate the efficacy of our approach in a number of benchmarks, and compare its performance to state-of-the-art methodologies.

The active monitoring of workload levels has been found to significantly reduce work-related stress. Heart rate and heart rate variability (HRV) measurements via photoplethysmography (PPG) sensors have shown a strong potential to accurately describe daily workload levels. However, due its complexity, HRV is commonly misunderstood and the associated measurements are rarely incorporated for workload monitoring in novel technological devices such as smartwatches and activity trackers. In this paper we explore the potential of consumer-grade smartwatches, equipped with PPG sensors, to assist in the active monitoring of workload during work hours. We develop a prototype that employs the SDNN index, a powerful HRV marker for cardiac resilience to differentiate between high and low workload levels along the work day, and presents feedback in glanceable form, by highlighting workload levels and physical activity over the past hour in 5-minutes blocks at the periphery of the smartwatch. A field study with 9 participants and 3 variations of our prototype attempts to quantify the impact of the HRV feedback over subjective and objective workload as well as users' engagement with the smartwatch. Results showed workload levels as inferred from the PPG sensor to positively correlate with self-reported workload and HRV feedback to result to lower levels of workload as compared to a conventional activity tracker. Moreover, users engaged more frequently with the smartwatch when HRV feedback was presented, than when only physical activity feedback was provided. The results suggest that HRV as inferred from PPG sensors in wearables can effectively be used to monitor workload levels during work hours.

Provenance data can be expressed as a graph with links informing who and which activities created, used and modified entities. The semantics of these links and domain specific reasoning can support the inference of additional information about the elements in the graph. If such elements include personal identifiers and/or personal identifiable information, then inferences may reveal unexpected links between elements, thus exposing personal data beyond an individual's intentions. Provenance graphs often entangle data relating to multiple individuals. It is therefore a challenge to protect personal data from unintended disclosure in provenance graphs. In this paper, we provide a Privacy Impact Assessment (PIA) template for identifying imminent privacy threats that arise from provenance graphs in an application-agnostic setting. The PIA template identifies privacy threats, lists potential countermeasures, helps to manage personal data protection risks, and maintains compliance with privacy data protection laws and regulations.

This paper proposes a new unsupervised satellite change detection method, which is robust to illumination changes. To achieve this, firstly, a preprocessing strategy is used to remove illumination artifacts and results in less false detection than traditional threshold-based algorithms. Then, we use the corrected input data to define a new fitness function based on the difference image. The purpose of using Self-Adaptive Hybrid Particle Swarm Optimization-Genetic Algorithm (SAPSOGA) is to combine two meta-heuristic optimization algorithms to search and find the feasible solution in the NP-hard change detection problem rapidly and efficiently. The hybrid algorithm is employed by letting the GA and PSO run simultaneously and similarities of GA and PSO have been considered to implement the algorithm, i.e. the population. In the SAPSOGA employed, in each iteration/generation the two population based algorithms share different amount of information or individual(s) between themselves. Thus, each algorithm informs each other about their best optimum results (fitness values and solution representations) which are obtained in their own population. The fitness function is minimized by using binary based SAPSOGA approach to produce binary change detection masks in each iteration to obtain the optimal change detection mask between two multi temporal multi spectral landsat images. The proposed approach effectively optimizes the change detection problem and finds the final change detection mask.

Software architecture is no more a mere system specification as resulting from the design phase, but it includes the process by which its specification was carried out. In this respect, design decisions in component-based software engineering play an important role: they are used to enhance the quality of the system, keep the current market level, keep partnership relationships, reduce costs, and so forth. For non trivial systems, a recurring situation is the selection of an asset origin, that is if going for in-house, outsourcing, open-source, or COTS, when in the need of a certain missing functionality. Usually, the decision making process follows a case-by-case approach, in which historical information is largely neglected. This solution avoids the overhead of keeping detailed documentation about past decisions, but hampers consistency among multiple, possibly related, decisions. The ORION project aims at developing a decision support framework in which historical decision information plays a pivotal role: it is used to analyse current decision scenarios, take well-founded decisions, and store the collected data for future exploitation. In this paper, we outline the potentials of such a knowledge repository, including the information it is intended to be stored in it, and when and how to retrieve it within a decision case. ORION

Testing is a key activity in industry to verify and validate products before they reach end customers. In hardwarein- the-loop system-level verification of automotive systems, testing is often performed using sequential execution of test scripts, each containing a mix of stimuli and assertions. In this paper, we propose and study an alternative approach for automated system-level testing automotive systems. In our approach, assertion-only test scripts and one (or several) stimulionly script(s), execute concurrently on the test driver. By separating the stimuli from the assertions, with each assertion independently determining when the system under test shall be verified, we seek to achieve three things: 1) tests that better represent real-world handling of the product, 2) reduced test execution time, and 3) increased defect detection. In addition to describing our proposed approach in detail, we provide experimental results from an industrial case study evaluating the approach in an automotive system test environment. PINT

The number of mobile devices is estimated to now exceed the world’s population, using more and more cloud services, and hence generating more and more traffic. Smartphones generate 95% of the total global handset traffic, and while approximately half of this traffic is sent to cellular networks, other handsets such as tablets are also using increasingly the cellular networks. This paper provides a closer look at the traffic generated on cellular networks by exploring billions of HTTP requests sent by millions of users to a nation-wide cellular network during 41 days. We confirm that - as in many other contexts - 20% of the users are responsible for more than 80% of the requests and provide a deeper analysis of the cellular network usage. Furthermore, we characterise the activity of users on their mobile device and which cloud services they use. For instance, almost 30% of the users use the cellular network frequently, mainly using search services and social networks, but 20% of their requests are sent to advertisement and tracking systems.

In many real-world scenarios, predictive models need to be interpretable, thus ruling out many machine learning techniques known to produce very accurate models, e.g., neural networks, support vector machines and all ensemble schemes. Most often, tree models or rule sets are used instead, typically resulting in significantly lower predictive performance. The over- all purpose of oracle coaching is to reduce this accuracy vs. comprehensibility trade-off by producing interpretable models optimized for the specific production set at hand. The method requires production set inputs to be present when generating the predictive model, a demand fulfilled in most, but not all, predic- tive modeling scenarios. In oracle coaching, a highly accurate, but opaque, model is first induced from the training data. This model (“the oracle”) is then used to label both the training instances and the production instances. Finally, interpretable models are trained using different combinations of the resulting data sets. In this paper, the oracle coaching produces regression trees, using neural networks and random forests as oracles. The experiments, using 32 publicly available data sets, show that the oracle coaching leads to significantly improved predictive performance, compared to standard induction. In addition, it is also shown that a highly accurate opaque model can be successfully used as a pre- processing step to reduce the noise typically present in data, even in situations where production inputs are not available. In fact, just augmenting or replacing training data with another copy of the training set, but with the predictions from the opaque model as targets, produced significantly more accurate and/or more compact regression trees. Sponsorship:This work was supported by the Swedish Foundation for StrategicResearch through the project High-Performance Data Mining for Drug EffectDetection (IIS11-0053), the Swedish Retail and Wholesale DevelopmentCouncil through the project Innovative Business Intelligence Tools (2013:5)and the Knowledge Foundation through the project Big Data Analytics byOnline Ensemble Learning (20120192).

The purpose of this paper is to propose and evaluate a method for reducing the inherent tendency of genetic programming to overfit small and noisy data sets. In addition, the use of different optimization criteria for symbolic regression is demonstrated. The key idea is to reduce the risk of overfitting noise in the training data by introducing an intermediate predictive model in the process. More specifically, instead of directly evolving a genetic regression model based on labeled training data, the first step is to generate a highly accurate ensemble model. Since ensembles are very robust, the resulting predictions will contain less noise than the original data set. In the second step, an interpretable model is evolved, using the ensemble predictions, instead of the true labels, as the target variable. Experiments on 175 sales forecasting data sets, from one of Sweden’s largest wholesale companies, show that the proposed technique obtained significantly better predictive performance, compared to both straightforward use of genetic programming and the standard M5P technique. Naturally, the level of improvement depends critically on the performance of the intermediate ensemble. Sponsorship:This work was supported by the Swedish Retail and Wholesale DevelopmentCouncil through the project Innovative Business Intelligence Tools (2013:5).

Voting advice applications (VAA) have become an increasingly popular feature of electoral campaigns. VAAs are online tools that use survey techniques to measure the degree to which the policy preferences of citizens match those of political parties or candidates. In some cases, such as The Netherlands, VAA's can attract millions of respondents providing an incredibly rich source of mass public opinion data. As a result political scientists have begun to exploit such datasets and this is fuelling a burgeoning literature on the topic. To date, however, there has been surprisingly little research on the cleaning techniques used to filter out the many rogues entries that are known to be present in VAA generated datasets. This paper presents the various methods used for cleaning VAA generated datasets that have been used for empirical research. Two main techniques are used based on item response timers and pattern recognition techniques. We show why cleaning matters and the problems that flow from not establishing rigorous cleaning techniques. The problem as such is not exclusive to VAA data but is common to all web based research involving self-administered surveys. To that end the techniques we present could be generalisable beyond the specific case of VAA-generated datasets.