Search

The Internet of Things (IoT) revolution has massively introduced sensor-rich devices to an ever growing landscape of smart environments. A key component in the IoT scenarios of the future is the requirement to utilize a shared database that allows all participants to operate collaboratively, transparently, immutably, correctly and with performance guarantees. Blockchain databases have been proposed by the community to alleviate these challenges, however existing blockchain architectures suffer from performance issues. In this vision paper we propose Triabase, a novel permissioned blockchain database system that carries out machine learning on the edge, abstracts machine learning models into primitive data blocks that are subsequently stored and retrieved from the blockchain. As such, it does not store detailed records on a medium, like blockchains, which is fundamentally very slow due to the expensive verification process. We lay out the primitive architectural blocks of our design, the requirements and the inherent challenges. Triabase employs technical novelties in respect to its consensus protocol, namely the notion of Proof-of-Federated-Learning (PoFL). The Triabase prototype system is implemented in the Hyperledger Fabric blockchain framework, upon which encouraging preliminary findings have been drawn.

Internet of Things (IoT) devices have penetrated massively into smart environments (e.g., smart-homes, smart-cars or more generally smart-anything). Besides data collection, many IoT devices also enable the execution of Rule Automation Workflows (RAW), which span from simple predicate statements to procedural workflows capturing a smart actuation pipeline. RAW aim to meet the convenience (comfort) level of users under specific conditions (e.g., raise room temperature to 22 C if cold), but unfortunately cannot express long-term objectives of users (e.g., consume less than 400 kWh in December). In this paper, we present an innovative system, coined IoT Meta-Control Firewall (IMCF), which internally deploys an AI-inspired Energy-Planner (EP) algorithm that exploits domain-specific operators to balance the trade-off between convenience and energy consumption in satisfying the RAW pipelines of users. IMCF filters the RAW pipelines in a way that these do not conflict with the long-term objectives of users (like a network firewall). Our experimental evaluation with extensive real traces from an apartment, a house, and campus dorms shows that IMCF achieves very high levels of user convenience while remaining within the target energy consumption budgets expressed by users.

Over the last years, Indoor Localization Systems (ILS) evolved, due to the inability of Global Positioning Systems (GPS) to localize in indoor environments. A variety of studies tackle indoor localization with technologies such as Bluetooth Beacons and RFID that require costly installation, or techniques such as Google Wi-Fi/Cell DB and fingerprinting that leverage from the already existing Wi-FI and telecommunication infrastructure. Additionally, recent studies attempt to solve the same problem using Bio-Inspired techniques, such as Artificial Neural Networks (ANNs) and Deep Neural Networks (DNN). In this paper, we introduce a Multi-Objective Optimization Radiomap Modelling (MOO-RM) based ILS. The MOO-RM ILS divides the dataset into clusters using a K-Means algorithm and trains ANN models on the data of each cluster. The resulting models are fed into a Multi-Objective Evolutionary Algorithm based on Decomposition (MOEA/D), which minimizes the required storage space and the localization error, simultaneously. Our experimental studies demonstrate the superiority of the proposed approach on real datasets of Wi-Fi traces with respect to various existing techniques.

The impressive growth of smartphone devices in combination with the rising ubiquity of using mobile platforms for sensitive applications such as Internet banking, have triggered a rapid increase in mobile malware. In recent literature, many studies examine Machine Learning techniques, as the most promising approach for mobile malware detection, without however quantifying the uncertainty involved in their detections. In this paper, we address this problem by proposing a machine learning dynamic analysis approach that provides provably valid confidence guarantees in each malware detection. Moreover the particular guarantees hold for both the malicious and benign classes independently and are unaffected by any bias in the data. The proposed approach is based on a novel machine learning framework, called Conformal Prediction, combined with a random forests classifier. We examine its performance on a large-scale dataset collected by installing 1866 malicious and 4816 benign applications on a real android device. We make this collection of dynamic analysis data available to the research community. The obtained experimental results demonstrate the empirical validity, usefulness and unbiased nature of the outputs produced by the proposed approach.

The evolution of ubiquitous smartphone devices has given rise to great opportunities with respect to the development of applications and services, many of which rely on sensitive user information. This explosion on the demand of smartphone applications has made them attractive to cybercriminals that develop mobile malware to gain access to sensitive data stored on smartphone devices. Traditional mobile malware detection approaches that can be roughly classified to signature-based and heuristic-based have essential drawbacks. The former rely on existing malware signatures and therefore cannot detect zero-day malware and the latter are prone to false positive detections. In this paper, we propose a heuristic-based approach that quantifies the uncertainty involved in each malware detection. In particular, our approach is based on a novel machine learning framework, called Conformal Prediction, for providing valid measures of confidence for each individual prediction, combined with a Multilayer Perceptron. Our experimental results on a real Android device demonstrate the empirical validity and both the informational and computational efficiency of our approach.

k Nearest Neighbor (kNN) search is one of the simplest non-parametric learning approaches, mainly used for classification and regression. kNN identifies the k nearest neighbors to a given node given a distance metric. A new challenging kNN task is to identify the k nearest neighbors for all nodes simultaneously; also known as All kNN (AkNN) search. Similarly, the Continuous All kNN (CAkNN) search answers an AkNN search in real-time on streaming data. Although such techniques find immediate application in computational intelligence tasks, among others, they have not been efficiently optimized to this date. We study specialized scalable solutions for AkNN and CAkNN processing as demanded by the volume–velocity-variety of data in the Big Data era. We present an algorithm, coined Proximity, which does not require any additional infrastructure or specialized hardware, and its efficiency is mainly attributed to our smart search space sharing technique. Its implementation is based on a novel data structure, coined k +-heap. Proximity, being parameter-free, performs efficiently in the face of high velocity and skewed data. In our analytical studies, we found that Proximity provides better time complexity compared to existing approaches and is very well suited for large scale scenarios.

The evolution of Smartphone devices with their powerful computing capabilities and their ever increasing number of sensors has recently introduced an unprecedented array of applications and games. The Smartphone users who are constantly moving and sensing are able to provide large amounts of opportunistic/participatory data that can contribute to complex and novel problem solving, unfolding in this way the full potential of crowdsourcing. Crowdsourced data can therefore be utilized for optimally modeling human-like behavior and improving the realizablity of AI gaming. In this study, we have developed an Augmented Reality First Person Shooter game, coined AR Shooter, that allows the crowd to constantly contribute their game play along with various spatio-temporal information. The crowdsourced data are used for modeling the human player’s behavior with Artificial Neural Networks. The resulting models are utilized back to the game’s environment through AI agents making it more realistic and challenging. Our experimental studies have shown that our AI agents are quite competitive, while being very difficult to distinguish from human players.

Computer technology seems to be very useful to support collaborative learning scenarios and thus Computer Supported Collaborative Learning. The main goal of this paper is to compare two of the most promising Collaborative Virtual Environment platforms, namely Croquet and Second Life, in terms of their potential to support collaborative e-learning scenarios among geographically separated users. Both platforms are examined by validating the platforms’ features, philosophy and policies against some basic design principles for collaborative virtual learning environments in order to better assess their design adequacy for online learning. Finally, this paper presents the user evaluation results, which was conducted by undergraduate students in order to assess the potential of both platforms for supporting collaborative e-learning scenarios.

In this paper we present a pedagogically-documented approach and model that facilitates the specialised and personalised e-training of professionals and end-users in application software products. The main focus of our methodology is on blended learning, authentic learning and competence based learning. Via blended learning, face to face as well as online teaching is used. Moreover, there is an e-training solution that can be used either as stand alone solution, for teaching a particular competence or can be used as a support to face to face teaching (that is revisit a competence taught in face to face). The e-Training component of the ELEVATE environment uses authentic learning by utilising “live” software systems. Competence based learning is used to identify the competence gap as well as tailor the training that the individual learner needs to attain the competence required.

This paper reviews and compares the most promising collaborative virtual environment platforms, which have been used or proposed for supporting educational activities in terms of their potential to support collaborative e-learning. The most promising environment according to the results of this review is Second Life. Second Life is further examined by validating the platform's features, philosophy and policies against some basic design principles for collaborative virtual learning environments in order to better assess its design adequacy for online learning. Furthermore, this paper presents the features that we have implemented within the Second Life platform, in order to facilitate the jigsaw collaborative e-learning scenario. Finally, we present a case study concerning the evaluation of Second Life by undergraduate students in order to assess its potential to support collaborative e-learning techniques.