Research Projects

Sponsored by the National Science Foundation (NSF)

and the Department of Defense (DoD)

Trajectory Privacy Preservation in Mobile Sensor Networks

Mentors: Dr. Pissinou, Dr. Kim

Mobile sensor networks are finding rapidly growing applications in a wide range of data collection applications, such as natural disaster forecasting, health condition monitoring, military reconnaissance, and traffic monitoring. Meanwhile, privacy is becoming an indispensable concern of these applications. This project targets an important but mostly untouched privacy issue: trajectory privacy preservation in mobile sensor networks. Mobile sensors, carried by users or vehicles, communicate with peer nodes and the base station through wireless media continuously. Therefore, trajectory information not only indicates the movement patterns of the sensors, but also reveals personal preferences and habits of users. Trajectory privacy invasion could lead to several types of risks for users, including personal safety, consuming profiling, and interference attacks. The research aims to answer the following interrelated questions: (1) what form of trajectory information could be invaded by attackers without accessing the database and how do they occur? (2) what methods can be used to protect trajectory privacy from both external and internal attacks undergoing data transmissions? (3) what are the critical factors for build a trajectory privacy attack-tolerant localization system? Answering these questions will result in distributed real-time trajectory privacy preservation techniques and a novel trajectory privacy-aware localization system framework. Privacy preservation mechanisms in sensor networks often require energy-efficient models and active contributions from multiple parties. Graph coloring theory and cognitive radio technology will be the basis of the energy-efficient models. Privacy mechanisms in the presence of selfish stakeholders will be also studied, notably by means of game theory. The outcome of this project will directly benefit mobile sensor users. Moreover, the success of this research will have a tremendous impact on advancing and inspiring the applications of mobile sensor networks.

In sum, ASSET students will (a) simulate a new routing protocol in network simulator OMNetPP; (b) participate in the development of the software to implement a novel privacy preserving routing protocol into mesmicsensor boards and gateways; (c) conduct static wireless sensor network experiments for trajectory data logging and network condition monitoring; (d) modify, calibrate and verify the sensor kits to properly operate with sensor mobility; (e) collect and analyze full-scale data in the experiments fields with 30-40 sensors and 10-20 gateways. The full-scale evidence acquisition includes source node trajectory data logging, node message dropping monitoring, receiving gateway trajectory data logging and network condition monitoring.

Digital Interventions for Reducing Social Networking Risks in Adolescents

Mentor: Dr. Carbunar

Adolescents are particularly exposed to risky behaviors in online social networks that include uploading and sharing sensitive personal information, clicking on suspicious links and sending/accepting friend invitations to/from users with no shared context. In addition to exposing them to cyber security and privacy threats, such behaviors have been shown to affect the safety of adolescents. In this project, the participants will develop digital intervention solutions to motivate, educate, support and engender safe social networking behaviors in adolescents. They will develop notifications and warnings for risky behaviors that are effective to provide a change in locus of attention and motivate users. The warnings leverage the unique combination of social, spatial and temporal dimensions provided by social networks. The participants will explore elements of "gamification" to devise reward mechanisms that convey and reinforce safety information in an effective and ongoing basis, and provide motivation for user participation. The participants will contribute to user studies to evaluate developed solutions.

Performance-Guaranteed Energy Conservation for Virtualized Data Centers by Stochastic Traffic Modeling

Mentor: Dr. Pan

Modern data centers contain thousands of servers and hundreds of switches. Their huge energy consumption has attracted significant attention. As servers become more energy efficient with various technologies, the data center network has been accounting for 20% or more of the energy consumed by the entire data center. On the other hand, studies on data center network traffic reveal that it has fluctuating patterns, and therefore demonstrates great potential for energy savings. In this project, we propose a suite of stochastic traffic modeling based optimization techniques for virtualized data centers to achieve energy conservation while provide performance guarantees. Specifically, the project will make research contributions along the following directions: stochastic traffic consolidation with performance guarantees, virtual switch assisted energy efficient routing, and joint host-network energy optimization for virtualized Data centers.

Network Function Virtualization in the Cloud

Mentor: Dr. Sun

Network function virtualization (NFV) is an emerging technology that seeks to improve the network agility by leveraging the massive capability of today�s cloud data centers to provide key network services such as intrusion detection and load balancing. Important benefits of NFV include reduced equipment costs and reduced power consumption through exploiting the economies of scale of cloud, and targeted service introduction based on geography or customer sets. However, as this technology is still in a very early stage, today there are few deployed NFV instances. This project seeks to understand the challenges in enabling NFV, by performing a case study of migrating several key network functions to the virtualized platform. Further, based on the experience gained through the case study, the project seeks to quantify the impact of virtualization in network performance and security, and explore ways to ensure the portability of the virtualized network functions between different cloud locations.

Trajectory Sensor Stream Cleaning

Mentor: Dr. Pissinou, Dr. Graham

While the emergence of sensors has fostered an increasing societal dependence on a reliable and continued availability of data, this dependence has been made much more critical by the new wireless paradigms where a plethora of mobile sensors communicate with each other to form mobile Wireless Sensor Networks (mWSNs). Sensors in mWSNs typically generate high volumes of data streams that can be used for applications that require a real-time response. It is clear, however, that sensors do not gather or transmit accurate data at all times. Interference and congestion alone minimize the quality of the data. Poor quality sensor data lowers the service quality of mWSN applications.

While researchers have attempted to improve the quality of the received sensor data, little work has been done in the field of mWSNs. In those previous works, the goal was to correct or "clean" the corrupted static data stored in the database. These methods required access to complete data sets that are stored in the databases and provide the comprehensive data cleaning solution, but they cannot provide a timely response to time-sensitive continuous queries inherent in mWSN. Therefore, these methods cannot be implemented to process data streams in sensor applications. Furthermore, the works assumed that the sensors were static and so the contextual relationships in time and space among them remain unchanged. Such techniques cannot be verbatim applied to mobile sensors which create dynamic contextual relationships among sensors.

Even stationary sensors with static surroundings may constantly experience changes in their operating environments as a result of changing queries. If sensor nodes are to meet their design requirements, they must be able to change their data gathering and cleaning operation to conform to the various contexts imposed by their positions, environments, malicious data streams and requests for data. Moreover, nodes must simultaneously, i.e., through rapid adaptation, handle multiple contexts. For example, those imposed by queries from different data sink nodes. To realize this goal, we need to explore how correlations between sensor streams affected by semantic trajectories, such as stop, go, visited or movement patterns, of collaborative neighboring sensors. Furthermore, we must determine how uncertainties in the received trajectory information influence the semantics and quality of sensor readings. Hence, our project involves the design, development and experimental demonstration of an online data stream cleaning methodology for mWSNs that incorporates context-awareness, adaptability, and dynamism.

In sum ASSET students will (a) design and develop a trajectory relationship model; (b) design and develop a dynamic context-aware annotation method of trajectory sensor streams and (c) design and develop a cleaning mechanism that tolerates the uncertainty of trajectory readings. Students will adopt an experimental approach in order to evaluate the degree to which incorporating concept of trajectory relationships and context awareness into the solution delivers accurate sensor data.

Making Cloud Computing Less "Thirsty"

Mentor: Dr. Ren

Data centers also not only energy hogs, but are also very "thirsty". A large data center may consume millions of gallons of cooling water each day; in addition, data centers also indirectly consume an enormous amount of water embedded in offsite electricity generation. As a result, water conservation is surfacing as a critical concern for data centers, amid the anticipation of surging water demand worldwide. Left unchecked, the growing water footprint of data centers can pose a severe threat to data center sustainability and may even handicap availability of services, especially for data centers in water-stressed areas. Existing mechanical solutions for conservation, such as using recycled/industry water and directly using outside cold air, are often costly and/or very limited by external factors such as locations, climate conditions, among others. As part of the integral efforts from both industry and academy to enable data center sustainability, this project uniquely integrates water footprint as an essential part of resource management in virtualized data centers.

Power Management in Multi-Tenant Data Centers

Mentor: Dr. Ren

Power-hungry data centers have been massively expanding in both number and scale, placing an increasing emphasis on optimizing data center power management. While the progress in data center energy efficiency is encouraging, the existing efforts have dominantly centered around owner-operated data centers (e.g., Microsoft). Another unique and integral segment of data center industry --- multi-tenant colocation data center, simply called "colocation", which is the physical home to many Internet and cloud services --- has not been well investigated, which, if still left unchecked, would become a major hurdle for sustainable growth of the digital economy. In sharp contrast with owner-operated data centers where operators have full control over both computing resources and facilities, colocation rents physical space to multiple tenants which individually control their own physical servers and power management, while the colocation operator is mainly responsible for facility support (e.g., providing reliable power and cooling). The uncoordinated power management, resulting from the colocation operator's lack of control over tenants' servers, invalidates many of the existing power management solutions for owner-operated data centers, thereby making colocations' operation highly inefficient. Thus, this project focuses on colocations and proposes to coordinate tenants' power management via market approaches.

Self-configuring, non-cooperative mobile sensors

Mentors: Dr. Iyengar, Dr. Pissinou

Future wireless communications will entail autonomous sensing devices to be interconnected in an ad hoc manner and without any underlying networked infrastructure, resulting in a temporary, on the fly, connection between typically autonomous sensing devices. This requires the participation of several sensor nodes, from multiple sensing systems and network domains with different objectives and preferences, where each device decides whether and to what extent it wishes to participate in the network. In the pursuit of their own interests, the participating sensing devices could therefore misbehave – either by being selfish or by being malicious. The main objective of our research is to use game theory to model and analyze strategic interactions in wireless, mobile sensor networks with non-cooperative, selfish and malicious users. The research will involve identifying the right equilibrium notion that captures the strategic interactions among wireless, mobile sensor nodes and employing game theoretic tools in the design of new wireless, mobile sensor architectures and protocols. It will focus on the study of various techniques for incorporating imperfect monitoring, imperfect information, and incomplete information into the game model for wireless, mobile sensor networks. Our aim is to design game theoretic algorithms for autonomous wireless, mobile sensor networks that achieve performance similar to those of cooperative networks with a central authority. We will adopt both an analytical and experimental approach to evaluate the results. Simulation is necessary to investigate certain behaviors in a truly heterogeneous mobile sensor network.

In sum, program participants will work on developing (1) new game-theoretic models to represent interaction among heterogeneous wireless and mobile sensor nodes. (2) new game theoretic tools for the analysis of complex interactions in non-cooperative wireless and mobile sensor domains and (3) new trust and privacy mechanisms using game for networks with autonomous mobile nodes without a central manager or a trusted authority.

Experiments in Adversarial Multi-Robot Patrolling

Mentor: Dr. Leonardo Bobadilla

In this project, we will investigate the problem of finding robust and easily implementable algorithms for multi-robot patrolling in adversarial settings. Multi-robot patrolling is the problem of visiting repeatedly a sequence of regions of interest in an environment. This fundamental problem has applications in surveillance, ecological monitoring, and search and rescue. The problem is challenging from both theoretical and practical perspectives. On the theoretical side, robots should be able to coordinate their decision-making by exchanging messages to achieve a global performance. On the practical side, robust behaviors must be obtained with limitations in sensing, computing, and communication. Several algorithms have been proposed for multi-robot patrolling. The problem is further classified as: 1) perimeter patrolling, concerned with monitoring the outer boundary of a polygonal region and 2) area patrolling, where the robots must ensure to visit periodically the interior of a polygon. An additional component to the problem involves an adversarial that has knowledge of the robot strategy and tries to penetrate the area under surveillance. Previous work contained a multi-robot perimeter patrolling algorithm in an adversarial setting was proposed. The algorithm had the following steps: 1) Discretization of the perimeter of the polygon in segments; 2) Simplification of the shape of the outer boundary to a circular graph where each vertex represents a segment of interest; 3) Definition of simplified motion models in the graph represented by a Markov chain parametrized by a value p; and finally 4) Finding the optimal value of p that maximizes the probability of penetration detection. The algorithm presented in previous work made some assumptions that may prevent its physical deployment: all the robots should 1) be placed uniformly spaced in the circular graph and 2) synchronize their motions to move at the same time and in the same direction. The goal of this project is to remove these assumptions to reduce the initial information and communication requirements of the problem. In sum, program participants will perform the following activities in the ten-week duration period: 1) Brief literature survey of techniques for adversarial multi-robot patrolling, elements of Markov chains, and Convex optimization; 2) Introduction to software tools for simulating mobile robots; 3) Software implementation of adversarial multi-patrolling algorithms; 4) Prototype physical deployment of the algorithms in an affordable mobile robotics platform; 5) Evaluation, write-up, and discussion of the results.

Implementation of Fast Local Computation Algorithms

Mentor: Dr. Xie

In this project, the faculty will study several recent published state-of-the art randomized algorithms and try to implement them on reasonably large inputs. The goal is to study the power and practicality of these algorithms.

Geometric methods for territory-aware information protection

Mentor: Dr. Zeng

For security, in some situations, information propagation through wireless is restricted in a specific region. This region may be highly irregular. Traditional methods are very time consuming. Then the challenge is how to efficiently propagate the information exactly within the user-prescribed region. This project aims to provide a novel geometric approach to this problem. It can be used for location-based mobile services and help design the propagation range for individual business. It can also be extended to privacy protection and national defense. The detailed tasks include: 1) Literature study; 2) Comparison with our method; 3) Applications design; and 4) Development. After the first stage, the framework can be extended to time-variant cases.

Linear-rank conjecture

Mentor: Dr. Ning

Consider a Boolean function f:{0,1}^n -->{0,1}. A natural measure of its complexity is how many "restrictions" one needs to impose on f to make it a constant function. In other words, what is the minimum number of linear constraints one has to impose on f to turn it into a trivial function. Let s denote the number of Fourier sparsity of f, i.e. number of non-zero Fourier coefficients of f. In a recent paper, we conjecture that polylog(s) restrictions are sufficient but the best bound known is only sqrt(s). The conjecture, if true, would confirm the famous log-rank conjecture for a large class of functions and also have interesting implications in the analysis of Boolean functions. This project is to test the conjecture over various known functions and random functions, and explore which type of functions require more restrictions.

Fraudulent Jobs in Crowdsourcing Sites

Mentor: Dr. Bogdan

Crowdsourcing sites such as Freelancer are the launching pad for a wide range of malicious behaviors in social networks. In this project we investigate the concept of fraudulent crowdsourcing jobs. We seek to understand and extract a model of interaction in such jobs between employers and workers. This project has the dual goal to identify fraudulently promoted products and to disrupt such interactions. The project will use web crawling, machine learning and game theory tools.

Mobile Health App for Seniors at Home

Mentor: Dr. Ruogu

In this project, we will develop a smart phone app to monitor and assist seniors at home alone for unusual health conditions and emergencies. Specifically we will use the smart phone's camera and gravity sensor to monitor if there is a fall or dizziness. The app will detect these emergency conditions and alerts the relatives or local hospitals.

Natural Language Processing and Computational Linguistics

Mentor: Dr. Mark Finlayson

Prof. Mark Finlayson's work intersects artificial intelligence, computational linguistics, cognitive science, and the digital humanities. His research focuses on the science of narrative (as a language object), including understanding the relationship between narrative, cognition, and culture, developing new computational methods and techniques for investigating questions related to language and narrative, and endowing machines with the ability to understand and use narratives for a variety of applications. Prof. Finlayson seeks highly motivated REUs and RETs to address important problems in this area, including building automatic natural language processing tools for extracting syntax and semantics; building user-facing tools for language corpus annotation; and collecting richly annotated corpora of stories. A selection of relevant papers, illustrating representative project topics, may be found at http://users.cis.fiu.edu/~markaf/

Game-based Statistics Education

Mentor: Dr. Prabakar

This project aims to stimulate interest in statistics education for middle school students. Basically, it facilitates students to create data sets with their own choices in games and then use these data sets for statistical analysis. This personal association with the data set motivates students to understand implications of statistics education. We propose a web based framework where students would sign-in and play a virtual game. Currently, we have chosen a shopping game where each student is given initial startup virtual cash amount. The student can purchase or sell item on the web through time-bound auction. The shopping behavior (selling items, purchasing items, profits, loses, etc.) of students will form basis for datasets. Datasets for a group of students can be used for statistical measures and the statistical results can be related to the shopping behaviors of the group members. This project involves system architecture, authentication and secure online interaction for students, seamless web design and interface, a database framework for data storage and retrieval, system monitoring/administration for project leaders and school teachers, game strategies, user interface designs, and development of statistical tools.