Principal component analysis and the residual error is an effective anomaly detection technique. In an environment where anomalies are present in the training set, the derived principal components can be skewed by the anomalies. A further aspect of anomaly detection is that data might be distributed across different nodes in a network and their communication to a centralized processing unit is prohibited due to communication cost. Current solutions to distributed anomaly detection rely on a hierarchical network infrastructure to aggregate data or models; however, in this environment, links close to the root of the tree become critical and congested. In this paper, an algorithm is proposed that is more robust in its derivation of the principal components of a training set containing anomalies. A distributed form of the algorithm is then derived where each node in a network can iterate towards the centralized solution by exchanging small matrices with neighboring nodes. Experimental evaluations on both synthetic and real-world data sets demonstrate the superior performance of the proposed approach in comparison to principal component analysis and alternative anomaly detection techniques. In addition, it is shown that in a variety of network infrastructures, the distributed form of the anomaly detection model is able to derive a close approximation of the centralized model.

We explore the idea of objects that, under some contextual conditions, become less “agreeable” to the user. Such an object may lower its performance level, become less usable, or start producing discouraging stimuli. Yet, it always remains fully functional and usable. We call such objects resistive. We describe three ways, in which making an object resistive may add value to the user. The resistive object may offer feedback on the user's past improper actions. It may encourage the user to improve her behavior at the moment. Or, it may offer the user an instant, discouraging evaluation of a bad operation on the object, which she is about to perform. In each case, resistivity is meant to help the user avoid undesirable behavior and choose a better course of action. In the paper we introduce the concepts of personal and community-wide resistive objects and provide a model of object resistivity. We exemplify the concepts with a number of ideas for resistive objects. 'otably, the examples may serve as an inspiration for quite a few experiments with end users. Finally, we hint at selected research issues.

Quantifying social interactions requires accurate, reliable and real-time recognition, of both users' interpersonal distance and relative orientation. DARSIS is the outcome of our research towards fulfilling these requirements based upon a non-intrusive opportunistic mechanism that solely relies on sensors and communication capabilities of off-the-shelf smartphones. We developed a novel hierarchical classifier for interpersonal distance estimation, produced by a substantive training set of Bluetooth Received Signal Strength Indicator (RSSI) and a concrete feature selection process. The presented relative orientation estimation mechanism addresses problems associated with lack of facing direction information in prior works, independently of the wearing position. In addition, DARSIS introduces a collaborative sensing scheme which allows on-the-fly exchange of facing direction information between users and facilitates the interpersonal distance recognition process by sharing RSSI values among devices. We show that the proposed interpersonal distance estimation models outperform state-of-the-art solutions and achieve up to 93.52% accuracy while DARSIS as a coherent system detects accurately 81.40% of the interactions in a real-world environment.

Kernel principal component analysis and the reconstruction error is an effective anomaly detection technique for non-linear data sets. In an environment where a phenomenon is generating data that is non-stationary, anomaly detection requires a recomputation of the kernel eigenspace in order to represent the current data distribution. Recomputation is a computationally complex operation and reducing computational complexity is therefore a key challenge. In this paper, we propose an algorithm that is able to accurately remove data from a kernel eigenspace without performing a batch recomputation. Coupled with a kernel eigenspace update, we demonstrate that our technique is able to remove and add data to a kernel eigenspace more accurately than existing techniques. An adaptive version determines an appropriately sized sliding window of data and when a model update is necessary. Experimental evaluations on both synthetic and real-world data sets demonstrate the superior performance of the proposed approach in comparison to alternative incremental KPCA approaches and alternative anomaly detection techniques.

The occurrence of short but recurrent opportunistic contacts between static infrastructure and mobile devices largely characterizes recent Internet of Things (IoT) applications for Smart Cities and Smart Buildings scenarios. In order to efficiently exploit such existing communication opportunities to access services, share and collect data, IoT applications cannot rely on standard discovery mechanisms that periodically probe the environment to discover resources. Discovery protocols resilient to different contact opportunities and able to optimize energy consumption when device contacts are not present are therefore required in order to avoid waste of energy especially in battery operated user devices. Additionally, such new discovery protocols need to optimize the time available for communication operations, while being able to adjust to application requirements, and balance energy consumption with respect to latency for contacts discovery. To this aim, we introduce CARD, a Context Aware Resource Discovery framework that leveraging Q-Learning techniques extends the functionalities of asynchronous neighbor discovery protocols, while being capable to reduce energy wastage and discovery latency. Simulation results show that CARD performs better than existing approaches and is resilient to a variety of real scenarios characterizing Smart Cities and Smart Building deployments.

A novel method for a mobile phone centric observation of a user's facing direction is presented. To estimate this direction, our proposed technique exploits the acceleration pattern that can be measured by a smartphone as the user is walking. For an accurate analysis of the acceleration pattern, the proposed approach benefits from a new trigonometric interpolation scheme. Our algorithm is independent of the initial orientation of the device and is adaptable to various wearing positions on a user's body, which gives the user a larger degree of freedom. A detailed description of the algorithm, which has been customized for a trouser pocket is presented. In addition, complementary hints for adaptation of the algorithm to other wearing positions along with an example of chest pocket position are provided. We have evaluated a prototype implementation of our algorithm on a smartphone, through several field experiments. It has been observed that our algorithm outperforms the conventional GPS and PCA-based techniques in terms of accuracy, reliability and energy consumption. The results also show that our approach has been able to handle the sudden variations of the user's direction. We have further incorporated our algorithm into a dead-reckoning application as an example of its real-world utility.

Anomaly detection in a Wireless Sensor Network is an important aspect of data analysis in order to facilitate intrusion and event detection. A key challenge is creating optimal classifiers constructed from training sets in which the anomaly rates are varying due to the existence of non-stationary distributions in the data. In this paper we propose an adaptive algorithm that can dynamically adjust the anomaly rate parameter, which can be represented by a model parameter of a one-class quarter-sphere support vector machine. This algorithm operates in an online, iterative manner producing an optimal model for a training set, which is presented sequentially. Our evaluations demonstrate that our algorithm is capable of constructing optimal models for a training set that minimizes the error rate on the classification set compared to a static model, where the anomaly rate is kept stationary.

This work addresses joint transceiver optimization for multiple-input, multiple-output (MIMO) systems. In practical systems the complete knowledge of channel state information (CSI) is hardly available at transmitter. To tackle this problem, we resort to the codebook approach to precoding design, where the receiver selects a precoding matrix from a finite set of pre-defined precoding matrices based on the instantaneous channel condition and delivers the index of the chosen precoding matrix to the transmitter via a bandwidth-constraint feedback channel. We show that, when the symbol constellation is improper, the joint codebook based precoding and equalization can be designed accordingly to achieve improved performance compared to the conventional system.

Emerging applications in Multihop Wireless Networks (MHWNs) require considerable processing power which often may be beyond the capability of individual nodes. Parallel processing provides a promising solution, which partitions a program into multiple small tasks and executes each task concurrently on independent nodes. However, multihop wireless communication is inevitable in such networks and it could have an adverse effect on distributed processing. In this paper, an adaptive intelligent task mapping together with a scheduling scheme based on a genetic algorithm is proposed to provide real-time guarantees. This solution enables efficient parallel processing in a way that only possible node collaborations with cost-effective communications are considered. Furthermore, in order to alleviate the power scarcity of MHWN, a hybrid fitness function is derived and embedded in the algorithm to extend the overall network lifetime via workload balancing among the collaborative nodes, while still ensuring the arbitrary application deadlines. Simulation results show significant performance improvement in various testing environments over existing mechanisms.

Clustering algorithms are widely used in Wireless Sensor Networks (WSNs), which however incurs significant energy consumption at Cluster Headers(CHs). Therefore, a re-clustering operation is typically used to balance the workload, where different CHs are selected and clusters are reorganized. However, a considerable number of control messages is initiated during this process which inevitably consumes on-board node energy. Hence, the question of how often the network should perform the re-clustering operation needs to be addressed. In this paper, a distributed re-clustering solution is proposed, which provides an energy-efficient re-clustering rate to conserve node energy while also equalizing the node energy consumption across the network. The proposed algorithm calculates the approximate amount of energy required to reorganize the clusters and to deliver the sensory data. By properly predicting the levels of the energy consumptions values, the appropriate frequency of performing the re-clustering operation can be determined, which reduces control message overhead. To the best of our knowledge, this is the first work that analytically analyzes the overhead in re-clustering a WSN, groups re-clustering rounds to reduce this overhead, and simultaneously equalizes node lifetimes. Performance results show that the proposed algorithm outperforms two other popular clustering algorithms in node energy conservation and node lifetime equalization.

In this paper we present the uDirect algorithm as a novel approach for mobile phone centric observation of a user's facing direction, through which the device and user orientations relative to earth coordinate are estimated. While the device orientation estimation is based on accelerometer and magnetometer measurements in standing mode, the unique behavior of measured acceleration during stance phase of a human's walking cycle is used for detecting user direction. Furthermore, the algorithm is independent of initial orientation of the device which gives the user higher space of freedom for long term observations. As the algorithm only relies on embedded accelerometer and magnetometer sensors of the mobile phone, it is not susceptible to shadowing effect as GPS. In addition, by performing independent estimations during each step of walking the model is robust to error accumulation. Evaluating the algorithm with 180 data samples from 10 participates has empirically confirmed the assumptions of our analytical model about the unique characteristics of the human stance phase for direction estimation. Moreover, our initial inspection has shown a system based on our algorithm outperforms conventional use of GPS and PCA analysis based techniques for walking distances more than 2 steps.

Emerging applications in Wireless Sensor Networks (WSNs) demand notable in-network processing capacities rather than simple data gathering and dissemination. Therefore, the performances of the network like latency and energy consumption are greatly affected by how the various application requirements are mapped to the processing nodes in the network. This paper investigates intelligent task mapping and scheduling techniques based on Genetic Algorithm (GA), and proposes a novel task allocation model and a multi-hop communication model to schedule both computation and communication activities in the WSN environment. A hybrid fitness function which balances the energy consumption among collaborative sensor nodes with application tolerable delays is presented to extend the network lifetime. Simulation results show that the proposed algorithm has a better capability of balancing the network lifetime with latency constraints in both homogeneous and heterogeneous networks.

Clustering algorithms have been extensively applied for energy conservation in wireless sensor networks (WSNs). In a clustered WSNs, cluster-heads (CHs) play an important role and drain energy more rapidly than other member nodes. Numerous mechanisms to optimize CH selection and cluster formation during the set-up phase have been proposed for extending the stable operation period of the network until any node depletes its energy. However, the existing mechanisms assume that the traffic load contributed by each node is the same, in other words, same amount of data are sent to CH from the member nodes during each scheduled round. This paper assumes the nodes contribute traffic load at different rates, and consequently proposes an energy-efficient clustering algorithm by considering both the residual node energy and the traffic load contribution of each node during the set-up phase. The proposed algorithm makes nodes with more residual energy and less traffic load contribution get more chances to become CHs. Furthermore, clusters are adaptively organized in a way that the deviation of ratio between the total cluster energy and the total cluster traffic load (ETRatio) is limited, in order to balance the energy usage among the clusters. Performance evaluation shows that the proposed algorithm extends the stable operation period of the network significantly.

Network scenarios beyond 3G assume the cooperation of operators with wireless access networks of different technologies in order to improve scalability and provide enhanced services to their mobile customers. While the selection of an optimised delivery path in such scenarios with multiple access networks is already a challenging task for unicast delivery, the problem becomes more severe for multicast services, where a potentially large group of heterogeneous receivers has to be served simultaneously via shared resources. In this paper we study the problem of selecting the optimal bearer paths for multicast services with groups of heterogeneous receivers in wireless networks with overlapping coverage. We propose an algorithm for bearer selection with different optimisation goals, demonstrating the existing tradeoff between user preference and resource efficiency.