Gautam S. Thakur

PROTECT: proximity-based trust-advisor using encounters for mobile societies

Many interactions between network users rely on trust, which is becoming particularly important given the security breaches in the Internet today. These problems are further exacerbated by the dynamics in wireless mobile networks. In this paper we address the issue of trust advisory and establishment in mobile networks, with application to ad hoc networks, including DTNs. We utilize encounters in mobile societies in novel ways, noticing that mobility provides opportunities to build proximity, location and similarity based trust. Four new trust advisor filters are introduced - including encounter frequency, duration, behavior vectors and behavior matrices - and evaluated over an extensive set of real-world traces collected from a major university. Two sets of statistical analyses are performed; the first examines the underlying encounter relationships in mobile societies, and the second evaluates DTN routing in mobile peer-to-peer networks using trust and selfishness models. We find that for the analyzed trace, trust filters are stable in terms of growth with time (3 filters have close to 90% overlap of users over a period of 9 weeks) and the results produced by different filters are noticeably different. In our analysis for trust and selfishness model, our trust filters largely undo the effect of selfishness on the unreachability in a network. Thus improving the connectivity in a network with selfish nodes. We hope that our initial promising results open the door for further research on proximity-based trust.

On the efficacy of mobility modeling for DTN evaluation: Analysis of encounter statistics and spatio-temporal preferences

In mobile networking, the main goal of mobility modeling and simulation is the ability to accurately reproduce effects of realistic mobility on the performance of networking protocols. In the areas of adhoc and delay tolerant networks (DTNs), recent work on mobility modeling focused on replicating metrics of encounter statistics and spatio-temporal preferences. No studies have been conducted, however, to show whether matching these metrics is sufficient to accurately reproduce DTN protocol performance. In this study, we address this specific problem, and attempt to show the sufficiency (or lack thereof) of existing encounter and mobility metrics in reproducing realistic effects of mobility on networking protocols. We first analyze the characteristics of two well-established mobility models; the random direction and the time-variant community (TVC) models, and study whether they capture encounter statistics and preference patterns observed in real-world traces. Second, we contrast the performance of epidemic routing in DTNs based on the mobility models, to that based on extensive mobility traces. We provide two main findings. First, careful parameterization of the models can indeed replicate the metrics in question (e.g., inter-encounter time distribution). Second, even carefully crafted mobility models surprisingly result in protocol performance that is dramatically different from the trace-driven performance. The difference in message delivery delays can reach 67%, while difference in reachability approaches 80%. Such findings strongly suggest the need to revisit mobility modeling. Furthermore, they clearly show the insufficiency of existing encounter and preference metrics as a measure of mobility model goodness. Systematically establishing a new set of meaningful mobility metrics should certainly be addressed in future works.

COBRA: A framework for the analysis of realistic mobility models

The future global Internet is going to have to cater to users that will be largely mobile. Mobility is one of the main factors affecting the design and performance of wireless networks. Mobility modeling has been an active field for the past decade, mostly focusing on matching a specific mobility or encounter metric with little focus on matching protocol performance. This study investigates the adequacy of existing mobility models in capturing various aspects of human mobility behavior (including communal behavior), as well as network protocol performance. This is achieved systematically through the introduction of a framework that includes a multi-dimensional mobility metric space. We then introduce COBRA, a new mobility model capable of spanning the mobility metric space to match realistic traces. A methodical analysis using a range of protocol (epidemic, spraywait, Prophet, and Bubble Rap) dependent and independent metrics (modularity) of various mobility models (SMOOTH and TVC) and traces (university campuses, and theme parks) is done. Our results indicate significant gaps in several metric dimensions between real traces and existing mobility models. Our findings show that COBRA matches communal aspect and realistic protocol performance, reducing the overhead gap (w.r.t existing models) from 80% to less than 12%, showing the efficacy of our framework.

PlanetSense: a real-time streaming and spatio-temporal analytics platform for gathering geo-spatial intelligence from open source data

Geospatial intelligence has traditionally relied on the use of archived and unvarying data for planning and exploration purposes. In consequence, the tools and methods that are architected to provide insight and generate projections only rely on such datasets. Albeit, if this approach has proven effective in several cases, such as land use identification and route mapping, it has severely restricted the ability of researchers to inculcate current information in their work. This approach is inadequate in scenarios requiring real-time information to act and to adjust in ever changing dynamic environments, such as evacuation and rescue missions. In this work, we propose PlanetSense, a platform for geospatial intelligence that is built to harness the existing power of archived data and add to that, the dynamics of real-time streams, seamlessly integrated with sophisticated data mining algorithms and analytics tools for generating operational intelligence on the fly. The platform has four main components -- i) GeoData Cloud -- a data architecture for storing and managing disparate datasets; ii) Mechanism to harvest real-time streaming data; iii) Data analytics framework; iv) Presentation and visualization through web interface and RESTful services. Using two case studies, we underpin the necessity of our platform in modeling ambient population and building occupancy at scale.

SHIELD: Social sensing and Help In Emergency using mobiLe Devices

School and College campuses face a perceived threat of violent crimes and require a realistic plan against unpredictable emergencies and disasters. Existing emergency systems (e.g., 911, campus-wide alerts) are quite useful, but provide delayed response (often tens of minutes) and do not utilize proximity or locality. There is a need to exploit proximitybased help for immediate response and to deter any crime. In this paper, we propose SHIELD, an on-campus emergency rescue and alert management service. It is a fully distributed infrastructureless platform based on proximity-enabled trust and cooperation. It relies on nearby localized responses sent using Bluetooth and/or WiFi to achieve minimal response time and maximal availability thereby augmenting the traditional notion of centralized emergency services. Analysis of campus crime statistics and WLAN traces surprisingly show a strong positive correlation (over 55%) between on-campus crime statistics and spatiotemporal density distribution of on-campus mobile users. This result is promising to develop a platform based on mutual trust and cooperation. Finally, we also show a prototype application to be used in such scenarios.

Modeling and characterization of vehicular density at scale

Future vehicular networks shall enable new classes of services and applications for car-to-car and car-to-roadside communication. The underlying vehicular mobility patterns significantly impact the operation and effectiveness of these services, and hence it is essential to model and characterize such patterns. In this paper, we examine the mobility of vehicles as a function of traffic density of more than 800 locations from six major metropolitan regions around the world. The traffic densities are generated from more than 25 million images and processed using background subtraction algorithm. The resulting vehicular density time series and distributions are then analyzed. It is found using the goodness-of-fit test that the vehicular density distribution follows heavy-tail distributions such as Log-gamma, Log-logistic, and Weibull in over 90% of these locations. Moreover, a heavy-tail gives rise to long-range dependence and self-similarity, which we studied by estimating the Hurst exponent (H). Our analysis based on seven different Hurst estimators signifies that the traffic patterns are stochastically self-similar (0.5 ≤ H ≤ 1.0). We believe this is an important finding, which will influence the design and deployment of the next generation vehicular network and also aid in the development of opportunistic communication services and applications for the vehicles. In addition, it shall provide a much needed input for the development of smart cities.

Analyzing Reliability of Virtual Machine Instances with Dynamic Pricing in the Public Cloud

This study presents reliability analysis of virtual machine instances in public cloud environments in the face of dynamic pricing. Different from traditional fixed pricing, dynamic pricing allows price to dynamically fluctuate over arbitrary period of time according to external factors such as supply and demand, excess capacity, etc. This pricing option introduces a new type of fault: virtual machine instances may be unexpectedly terminated due to conflicts in the original bid price and the current offered price. This new class of fault under dynamic pricing may be more dominant than traditional faults in cloud computing environments, where resource availability associated with traditional faults is often above 99.9%. To address and understand this new type of fault, we translated two classic reliability metrics, mean time between failures and availability, to the Amazon Web Services spot market using historical price data. We also validated our findings by submitting actual bids in the spot market. We found that overall, our historical analysis and experimental validation lined up well. Based upon these experimental results, we also provided suggestions and techniques to maximize overall reliability of virtual machine instances under dynamic pricing.

Proximity based trust-advisor using encounters for mobile societies: Analysis of four filters

Many interactions between network users rely on trust, which is becoming particularly important given the security breaches in the Internet today. These problems are further exacerbated by the dynamics in wireless mobile networks. In this paper, we address the issue of trust advisory and its establishment in mobile networks, with application to ad hoc networks, including delay tolerant (DTNs). We utilize encounters in mobile societies in novel ways, noticing that mobility provides opportunities to build proximity, location, and similarity based trust. Four new trust advisor filters are introduced – including encounter frequency, duration, behavior vectors, and behavior matrices. The filters are evaluated over an extensive set of real-world traces collected from a major university. Two sets of statistical analyses are performed; the first examines the underlying encounter relationships in mobile societies, and the second evaluates DTN routing in mobile peer-to-peer networks using trust and selfishness models. We find that for the analyzed trace, trust filters are stable in terms of growth with time (three filters have close to 90% overlap of users over a period of 9 weeks) and the results produced by different filters are noticeably different. In our analysis for trust and selfishness model, our trust filters largely undo the effect of selfishness on the unreachability in a network. Thus improving the connectivity in a network with selfish nodes. We hope that our initial promising results open the door for further research on proximity-based trust. Copyright

Application of Social Media Data to High-Resolution Mapping of a Special Event Population

Society’s increasing participation in social media provides access to new sources of near-real-time data that reflect our activities in space and in time. The ability for users to capture and express their geolocations through their phones’ global positioning system (GPS), or through a particular location’s hashtag or Facebook page, provides an opportunity for modeling spatiotemporal population dynamics. One illustrative application is the modeling of dynamic populations associated with special events such as sporting events. To demonstrate, Twitter posts and Facebook check-ins were collected across a 24 h period for several football game days at the University of Tennessee, Knoxville, during the 2013 season. Population distributions for game hours and nongame hours of a typical game day were modeled at a high spatial resolution using the spatiotemporal distributions of the social media data.

Facility detection and popularity assessment from text classification of social media and crowdsourced data

Advances in technology have continually progressed our understanding of where people are, how they use the environment around them, and why they are at their current location. Having a better knowledge of when various locations become popular through space and time could have large impacts on research fields like urban dynamics and energy consumption. In this paper, we discuss the ability to identify and locate various facility types (e.g. restaurant, airport, stadiums) using social media, and assess methods in determining when these facilities become popular over time. We use standard natural language processing tools and machine learning classifiers to interpret geotagged Twitter text and determine if a user is seemingly at a location of interest when the tweet was sent. On average our classifiers are approximately 85% accurate varying across multiple facility types, with a peak precision of 98%. By using these standard methods to classify unstructured text, geotagged social media data can be an extremely useful tool to better understanding the composition of places and how and when people use them.