Data quality is a challenging problem in many real world application domains. While a lot of attention has been given to detect anomalies for data at rest, detecting anomalies for streaming applications still largely remains an open problem. For applications involving several data streams, the challenge of detecting anomalies has become harder over time, as data can dynamically evolve in subtle ways following changes in the underlying infrastructure. In this paper, we describe and empirically evaluate an online anomaly detection pipeline that satisfies two key conditions: generality and scalability. Our technique works on numerical data as well as on categorical data and makes no assumption on the underlying data distributions. We implement two metrics, relative entropy and Pearson correlation, to dynamically detect anomalies. The two metrics we use provide an efficient and effective detection of anomalies over high velocity streams of events. In the following, we describe the design and implementation of our approach in a Big Data scenario using state-of-the-art streaming components. Specifically, we build on Kafka queues and Spark Streaming for realizing our approach while satisfying the generality and scalability requirements given above. We show how a combination of the two metrics we put forward can be applied to detect several types of anomalies - like infrastructure failures, hardware misconfiguration or user-driven anomalies - in large-scale telecommunication networks. We also discuss the merits and limitations of the resulting architecture and empirically evaluate its scalability on a real deployment over live streams capturing events from millions of mobile devices.

We propose a new class of applications for Intelligent Transportation Systems (ITSs), called collaborative transportation applications that aim at solving transportation problems such as congestion and parking. Specifically, we define two applications: SmartPark and SmartRide that leverage short-range wireless communication. We quantify the potential benefits these collaborative transportation applications can offer to an individual and to the public. To this extent, we conduct both the realistic simulations and the analysis of the performance of a taxi cab fleet from San Francisco. Our analysis shows that both collaborative transportation applications can provide with significant savings in travel times, fuel consumptions, etc. Finally, we discuss the functional requirements of collaborative transportation applications and we present the challenges that these applications are facing.

We describe our design of a geocast service for mobile partitioned networks (MPNs). We focus mainly on minimizing the delivery latency. Our approach exploits the time-stability of the collective mobility pattern. In MPNs, in contrast to MANETs, the end-to-end path is frequently not available. Thus, communication in such networks becomes problematic. To overcome this difficulty, researchers propose a solution in which the nodepsilas mobility is exploited. This paradigm is often called mobility-assisted forwarding. In order to design routing protocols for MPNs, researchers study key mobility metrics, for example the inter-contact times between nodes. Based on the analysis of a real-life mobility trace, we show that the inter-contact time distribution is spatially dependent. This is a result of spatially heterogeneous mobility pattern that appears to be stable in time. We demonstrate that any georouting protocol designed to work in MPNs can benefit from knowing such an underlying mobility pattern. We propose an abstraction called mobility map that represents the collective mobility pattern. We also present how mobility maps can be used for georouting in MPNs and we also show a simple mechanism for the collaborative discovery of mobility maps. Finally, we propose a geocast protocol for MPNs - GeoMobCast - that explicitly uses mobility maps and is designed to minimize the expected message delay while maximizing the message delivery. We empirically evaluate the protocol by using simulations and we observe the improved performance, compared to other approaches.

Mobile wireless ad hoc and sensor networks can be permanently partitioned in many interesting scenarios. This implies that instantaneous end-to-end routes do not exist. Nevertheless, when nodes are mobile, it is possible to forward messages to their destinations through mobility. We observe that in many practical settings, spatial node distributions are very heterogeneous and possess concentration points of high node density. The locations of these concentration points and the flow of nodes between them tend to be stable over time. This motivates a novel mobility model, where nodes move randomly between stable islands of connectivity, where they are likely to encounter other nodes, while connectivity is very limited outside these islands. Our goal is to exploit such a stable topology of concentration points by developing algorithms that allow nodes to collaborate to discover this topology and to use it for efficient mobility forwarding. We achieve this without any external signals to nodes, such as geographic positions or fixed beacons; instead, we rely only on the evolution of the set of neighbors of each node. We propose an algorithm for this collaborative graph discovery problem and show that the inferred topology can greatly improve the efficiency of mobility forwarding. Using both synthetic and data-driven mobility models we show through simulations that our approach achieves end-to-end delays comparable to those of epidemic approaches, while requiring a significantly lower transmission overhead