Search

The increasing ubiquity of the mobile phone is creating many opportunities for personal context sensing, and will result in massive databases of individuals' sensitive information incorporating locations, movements, images, text annotations, and even health data. In existing system architectures, users upload their raw (unprocessed or filtered) data streams directly to content-service providers and have little control over their data once they "opt-in". We present Personal Data Vaults (PDVs), a privacy architecture in which individuals retain ownership of their data. Data are routinely filtered before being shared with content-service providers, and users or data custodian services can participate in making controlled data-sharing decisions. Introducing a PDV gives users flexible and granular access control over data. To reduce the burden on users and improve usability, we explore three mechanisms for managing data policies: Granular ACL, Trace-audit and Rule Recommender. We have implemented a proof-of-concept PDV and evaluated it using real data traces collected from two personal participatory sensing applications.

Advances in mobile phone technology have allowed phones to become a convenient platform for real-time assessment of a participants health and behavior. AndWellness, a personal data collection system, uses mobile phones to collect and analyze data from both active, triggered user experience samples and passive logging of onboard environmental sensors. The system includes an application that runs on Android based mobile phones, server software that manages deployments and acts as a central repository for data, and a dashboard front end for both participants and researchers to visualize incoming data in real-time. Our system has gone through testing by researchers in preparation for deployment with participants, and we describe an initial qualitative study plus several planned future studies to demonstrate how our system can be used to better understand a user's health related habits and observations.

The rapid adoption of mobile devices that are able to capture and transmit a wide variety of sensing modalities (media and location) has enabled a new data collection paradigm - participatory sensing. Participatory sensing initiatives organize individuals to gather sensed information using mobile devices through cooperative data collection. A major factor in the success of these data collection projects is sustained, high quality participation. However, since data capture requires a time and energy commitment from individuals, incentives are often introduced to motivate participants. In this work, we investigate the use of micro-payments as an incentive model. We define a set of metrics that can be used to evaluate the effectiveness of incentives and report on findings from a pilot study using various micro-payment schemes in a university campus sustainability initiative.

PEIR, the Personal Environmental Impact Report, is a participatory sensing application that uses location data sampled from everyday mobile phones to calculate personalized estimates of environmental impact and exposure. It is an example of an important class of emerging mobile systems that combine the distributed processing capacity of the web with the personal reach of mobile technology. This paper documents and evaluates the running PEIR system, which includes mobile handset based GPS location data collection, and server-side processing stages such as HMM-based activity classification (to determine transportation mode); automatic location data segmentation into "trips''; lookup of traffic, weather, and other context data needed by the models; and environmental impact and exposure calculation using efficient implementations of established models. Additionally, we describe the user interface components of PEIR and present usage statistics from a two month snapshot of system use. The paper also outlines new algorithmic components developed based on experience with the system and undergoing testing for integration into PEIR, including: new map-matching and GSM-augmented activity classification techniques, and a selective hiding mechanism that generates believable proxy traces for times a user does not want their real location revealed.

Sensor networks have been widely used to collect data about the environment. When analyzing data from these systems, people tend to ask exploratory questions---they want to find subsets of data, namely signal, reflecting some characteristics of the environment. In this paper, we study the problem of searching for drops in sensor data. Specifically, the search is to find periods in history when a certain amount of drop over a threshold occurs in data within a time span. We propose a framework, SegDiff, for extracting features, compressing them, and transforming the search into standard database queries. Approximate results are returned from the framework with the guarantee that no true events are missed and false positives are within a user specified tolerance. The framework efficiently utilizes space and provides fast response to users' search. Experimental results with real world data demonstrate the efficiency of our framework with respect to feature size and search time.

There exist many natural phenomena where direct measurement is either impossible or extremely invasive. To obtain approximate measurements of these phenomena we can build prediction models based on other sensing modalities such as features extracted from data collected by an imager. These models are derived from controlled experiments performed under laboratory conditions, and can then be applied to the associated event in nature. In this paper we explore various different methods for generating such models and discuss their accuracy, robustness, and computational complexity. Given sufficiently computationally simple models, we can eventually push their computation down towards the sensor nodes themselves to reduce the amount of data required to both flow through the network and be stored in a database. The addition of these models turn in-situ imagers into powerful biological sensors, and image databases into useful records of biological activity.

Imagers are an increasingly significant source of sensory observations about human activity and the urban environment. ImageScape is a software tool for processing, clustering, and browsing large sets of images. Implemented as a set of web services with an Adobe Flash-based user interface, it supports clustering by both image features and context tags, as well as re-tagging of images in the user interface. Though expected to be useful in many applications, ImageScape was designed as an analysis component of DietSense, a software system under development at UCLA to support (1) the use of mobile devices for automatic multimedia documentation of dietary choices with just-in-time annotation, (2) efficient post facto review of captured media by participants and researchers, and (3) easy authoring and dissemination of the automatic data collection protocols. A pilot study, in which participants ran software that enabled their phones to autonomously capture images of their plates during mealtime, was conducted using an early prototype of the DietSense system, and the resulting image set used in the creation of ImageScape. ImageScape will support two kinds of users within the DietSense application: The participants in dietary studies will have the ability to easily audit their images, while the recipients of the images, health care professionals managing studies and performing analysis, will be able to rapidly browse and annotate large sets of images.

Monitoring of environmental phenomena with embedded networked sensing confronts the challenges of both unpredictable variability in the spatial distribution of phenomena, coupled with demands for a high spatial sampling rate in three dimensions. For example, low distortion mapping of critical solar radiation properties in forest environments may require two-dimensional spatial sampling rates of greater than 10 samples/m2 over transects exceeding 1000 m2. Clearly, adequate sampling coverage of such a transect requires an impractically large number of sensing nodes. This paper describes a new approach where the deployment of a combination of autonomous-articulated and static sensor nodes enables sufficient spatiotemporal sampling densityo ver large transects to meet a general set of environmental mapping demands.To achieve this we have developed an embedded networked sensor architecture that merges sensing and articulation with adaptive algorithms that are responsive to both variabilityin environmental phenomena discovered bythe mobile sensors and to discrete events discovered byst atic sensors. We begin byde scribing the class of important driving applications, the statistical foundations for this new approach, and task allocation. We then describe our experimental implementation of adaptive, event aware, exploration algorithms, which exploit our wireless, articulated sensors operating with deterministic motion over large areas. Results of experimental measurements and the relationship among sampling methods, event arrival rate, and sampling performance are presented.

As part of the process of delivering content, devices like proxies and gateways log valuable information about the activities and navigation patterns of users on the Web. In this study, we consider how this navigation data can be used to improve Web search. A query posted to a search engine together with the set of pages accessed during a search task is known as a search session. We develop a mixture model for the observed set of search sessions, and propose variants of the classical EM algorithm for training. The model itself yields a type of navigation-based query clustering. By implicitly borrowing strength between related queries, the mixture formulation allows us to identify the "highly relevant" URLs for each query cluster. Next, we explore methods for incorporating existing labeled data (the Yahoo! directory, for example) to speed convergence and help resolve low-traffic clusters. Finally, the mixture formulation also provides for a simple, hierarchical display of search results based on the query clusters. The effectiveness of our approach is evaluated using proxy access logs for the outgoing Lucent proxy.

Abstracf- An approach to specification of requirements and verification of design for real-time systems is presented. A system is defined by a conventional mathematical model for a dynamic system where application specific states denote functions of real time. Specifications are formulas in duration calculus, a realtime interval logic, where predicates define durations of states. Requirements define safety and functionality constraints on the system or a component. A top-level design is given by a control law: a predicate that defines an automaton controlling the transition between phases of operation. Each phase maintains certain relations among the system states; this is analogous to the control functions known from conventional control theory. The top-level design is decomposed into an architecture for a distributed system with specifications for sensor, actuator, and program components. Programs control the distributed computation through synchronous events. Sensors and actuators relate events with system states. Verification is a deduction showing that a design implies requirements.