Department of Computer Science and Electrical Engineering

Ph.D. Dissertation Defense

Joint Inference for Extracting Soft Biometric
Text Descriptorsfrom Patient Triage Images

Niyati Chhaya

10:00am Friday, 24 August 2012, ITE 325b, UMBC

Disaster events can result in mass casualties and missing persons, giving rise to a need to provide information about victims to the public. This can be achieved by digitally documenting information available at emergency medical care centers in the form of pictures. The images and other identifying information, such as fingerprints, cannot be broadcast due to privacy concerns, leading to a need to extract appearance-related non-unique features from this data to facilitate locating missing persons. Using humans and machines to compare images is not feasible due to the scale of the situation and the nature (presence of blood and debris) of the images. Extracting a soft biometric text descriptor (text labels describing different soft biometric features) makes it possible to organize information about individuals from these images in a searchable format without revealing the person's identity. The main aim of this thesis is to extract soft biometric features from person images to label appearance-related information and make it available as a text descriptor.

We begin by presenting soft biometric feature detectors for patient images that include an ensemble-based face detection algorithm, template-based eye detection, and eyeglasses, hair color, and skin color detection. We also present a facial hair detector that uses a combination of face and hair information. The feature detection results indicate a need to combine and exploit feature relationships for better performance. We propose a novel probabilistic graphical model that consists of different feature detectors and exploits relationships between these features using a message-passing inference algorithm to build a coherent text descriptor. Further, to understand the utility and the nature of the text descriptors, we present a study based on human descriptions that aims at extracting order and structure information about the features.

We evaluate the performance of individual feature detectors for standard and triage images and establish the challenges posed by the latter. Further, our text analysis shows extreme variability in human descriptions. However, we succeed in extracting some insights about the order of a natural text description. Through our evaluations of the graphical model, we show that for different feature detectors, datasets, and graph sizes the graphical model helps improve the accuracy of the text output. We also show that the performance of the graphical model depends on the individual nodes (feature detectors) and that the model can be used to improve the performance of the individual feature detectors. This thesis illustrates the whole process from images to text descriptors while evaluating components as we proceed. This work presents an approach to extract text labels from images using computer vision, a probabilistic graphical model, and natural language processing techniques.

Committee: Drs. Dr. Tim Oates (Chair), Marie desJardins, Tim Finin, Glenn Pearson and Jesus Caban

CSEE Professor Dr. Alan Sherman (PI) and Dr. Rick Forno (Co-PI), Graduate Program Director, Cybersecurity have received an NSF grant of $2.5 million over five years to fund 22 students studying Information Assurance (IA) and Cybersecurity. The scholarships are part of the Federal Cyber Scholarship for Service (SFS) program.

Future scholars will come from UMBC’s undergraduate, graduate, and professional studies programs. In addition to a generous full-ride scholarship, scholars are granted an annual stipend during their last two or three years.  The yearly stipends are $20,000 for underdergraduate, $25,000 for M.S. and $30,000 for PhD students.  Students will also engage in paid summer internships with federal government agencies and have opportunities to carry out mentored research projects at UMBC and its cyber partners from government and industry.

“Our goal is to encourage students to pursue cybersecurity education at UMBC and then move into careers that can build upon their education,” explains Dr. Forno. “The SFS program is one way to do that, with the outcome being to provide educated, qualified and highly competent cybersecurity practitioners to the federal government.”

The scholarships will be coordinated through UMBC’s Center for Information Security and Assurance (CISA), and both CISA and UMBC’s Cyber Defense Lab (CDL) will play a key role in the overall academic experience of the SFS scholars.

Along with the monetary award come a few scholarship program requirements. During their scholarship period (typically two years), students are required to intern at a federal organization. After graduation, scholars must serve at a Federal agency in an information assurance position for one to two years, depending on their level of support.

During the upcoming 2012-2013 school year, the program will support two Masters students and two Master in Professional Studies students, though in the future, one B.S., two M.S., two M.P.S., and one Ph.D. students will be supported annually.

Degree-seeking students in Computer Science, Computer Engineering, or any related field, including Electrical Engineering, Math, Physics, Information Systems, Public Policy, or Education, are eligible. “We are interested in fostering a diverse group of SFS scholars,” says Forno. “To that end, we intend to work closely with groups like CWIT and the Meyerhoff Scholars (among others) to help in our outreach and recruiting efforts in this critical academic discipline.” To learn more about applying for the scholarship, visit the CISA scholarship oppotunities page.

The grant will also fund a new one-day annual regional workshop that will focus on innovations for K-12, undergraduate, and graduate cybersecurity education. The goal is to develop "innovative, relevant, meaningful, and real-world oriented pedagogical materials that can be applied across the academic spectrum in conducting cybersecurity education."

The scholarship program, coupled with UMBC’s location at the “epicenter of cybersecurity,” will further mark UMBC as a pioneer in the field. “We all know that cybersecurity is a hot field and of critical national interest," says Forno. "UMBC’s strength in Computer Science, Cybersecurity, and related STEM fields, combined with our reputation for producing qualified technologitsts makes our participation in SFS very attractive both to the school and our current and future students.”

Ph.D. Dissertation Defense

A Trust and Reputation Mechanism Through
Behavioral Modeling of Reviewers

Yasaman Haghpanah

11:00am Tuesday 21 August 2012, ITE 325b, UMBC

Trust and reputation have become important topics in various domains, such as online markets, supply chain management, auctions, social networks, and e-commerce applications. With the significant increase in transactions with people and organizations especially in online markets, people need to interact with strangers with whom they have little or no previous interactions. Reputation information as a form of world of mouth in auctions and supply chain management and as a form of provided reviews and ratings on online websites are two different sources for modeling trust and reputation in order to mitigate the risk of not knowing a stranger before actually start interacting with that stranger.

In providing reputation information, people can have different behavior, such as being biased based on incentives or they can have different preferences and viewpoints. In this dissertation, I introduce a novel trust and reputation mechanism that models and learns a reputation provider’s behavior based on probability theory. This learned behavior is then used to re-interpret the reputation information, thus making use of the entire reputation data effectively, even if they are biased or based on personal viewpoints and preferences.I show the importance of learning the behavior of reputation providers using different patterns of being biased or having different preferences and satisfaction thresholds in three different settings of game-theory, an online rating website, and an online marketplace. My results show that learning the behavior of reputation providers in all three above settings helps individuals to more effectively aggregate and adjust reputation information in order to make decisions, thereby increasing their satisfaction and overall payoffs in their interactions.

Committee: Drs. Marie desJardins (Chair), Tim Oates, Tim Finin, Wolfgang Ketter and David Aha

Congratulations to Xianshu Zhu (CS, Ph.D.), who won the best student paper award at the 13^th IEEE International Conference on Information Reuse and Integration.

The conference, which took place last weekend in Las Vegas, is a forum to discuss how we can maximize the reuse of information with the creation of simple, rich, and reusable knowledge representations.

Xianshu’s paper, “Finding Story Chain in Newswire Articles”, seeks to do just that through a proposed walk based algorithm that can be used to find news story chains on the internet. The algorithm sifts through the huge and hard to navigate database of news articles on the internet, resulting in a simple way to track the evolution of a news story, without redundant articles.

Xianshu’s areas of study include data mining, text mining, information retrieval. The paper is closely tied to the research she’s doing for her Ph.D. thesis (advised by Professor Tim Oates), which she plans on finishing up in May 2013.

PhD Dissertation Preliminary Examination

On Optimizing Contrast Quality and Acquisition Time of
SSFP-Sequence-Based techniques for Structural and Functional
MR Imaging via Extended Phase Graph (EPG) Analysis

Albert Kir

1:00pm Wednesday, 15 August 2012, ITE 325b

The extended phase graph (EPG) formalism provides an excellent description of the magnetization evolution (considering the magnetization as a vector quantity in space and its change through time) for steady-state free precession techniques (sequences). This dissertation demonstrates its accuracy by applying it to analyze/optimize structural and functional magnetic resonance imaging. An optimization framework for a structural imaging technique, Magnetization-Prepared RApid Gradient-Echo (MP-RAGE), based on the EPG algorithm is established for obtaining optimal images with respect to image signal strength, contrast, and acquisition time. In addition, a functional imaging technique, True Fast Imaging with Steady-state Precession (TruFISP) with improved quantitative sensitivity is developed using EPG analysis.

Committee: Drs. Joel M. Morris (chair and co-advisor), Alan McMillan (co-advisor), Rao Gullapalli (co-advosor), Janet Rutledge and Tulay Adali

PhD Preliminary Examination

Design and Analysis of Underwater
Acoustic Networks with Reflected Links

Lloyd Emokpae

2:00pm Monday, 13 August 2012, ITE 325b, UMBC

Underwater acoustic networks (UWANs) have applications in environmental state monitoring, oceanic profile measurements, leak detection in oil fields, distributed surveillance, and navigation. For these applications, a set of nodes are employed to collaboratively monitor an area of interest and track certain events or phenomena. In addition, it is common to find autonomous underwater vehicles (AUVs) acting as mobile sensor nodes that perform search-and-rescue missions, reconnaissance in combat zones and coastal patrol. These AUVs are to work cooperatively to achieve a desired goal and thus need to be able to, in an ad-hoc manner, establish and sustain communication links in order to ensure some desired level of quality of service. Therefore, each node is required to adapt to environmental changes and be able to overcome broken communication links caused by external noise affecting the communication channel and due to node mobility. In addition, since radio waves tend to get absorbed in the water, it is common for most underwater applications to rely on acoustic rather than radio channels for long range communications. However, acoustic channels pose multiple challenging issues, most notably the high transmission delay due to slow signal propagation and the limited channel bandwidth due to high frequency attenuation. Furthermore, the inhomogeneous property of the water medium affects the sound speed profile while the signal surface and bottom reflections leads to multipath effects.

In this dissertation proposal, we address these networking challenges by developing protocols that take into consideration the underwater physical layer dynamics. We begin by introducing a novel surface-based reflection scheme (SBR) which takes advantage of the multipath effects of the acoustic channel. SBR works by using reflections from the water surface, and bottom, to establish non-line-of-sight communication links. SBR makes it possible to incorporate both line-of-sight and non-line-of-sight links by utilizing directional antennas, which will boost the signal-to-noise ratio at the receiver while promoting non-line-of-sight usage. Furthermore, we employ a directional underwater acoustic antenna which is composed of an array of hydrophones that can be summed up at various phases and amplitudes resulting in a beam-former. Networking protocols have been developed for node discovery, localization, medium access arbitration and geographical routing to exploit the SBR communication model.

To complete the dissertation, we plan to first extend the SBR scheme to be robust to Doppler frequency shifts. In our original implementation of SBR, we assume little to no drift between communicating nodes which might be acceptable for some applications. However, when dealing with AUVs, the effects of Doppler could affect the proposed channel impulse response recovery process if it is not handled properly. Thus, we plan on fine-tuning the SBR recovery mechanism to adjust for Doppler frequency shifts in real-time. Secondly, we plan on extending the proposed medium access arbitration protocol to account for varying water surface and bottom roughness which we assumed to be flat in our preliminary simulations. Thirdly, we intend to extend the proposed routing algorithm to account for multiple metrics, e.g., network delay, error rate and energy. Finally, we plan on building an acoustic modem prototype which utilizes a multimodal directional transducer to validate the SBR scheme. We also plan on validating the SBR scheme on the proposed acoustic modem using a water tank.

Committee: Drs. Mohamed Younis (Chair), John Pinkston, Tulay Adali and Ryan Robucci

MS Defense

Approach to Unwrap a 3D Fingerprint to a 2D Equivalent

Ravikiran Dighade

10:00am Thursday, 2 August 2012, ITE 352

Fingerprints are the most widely used biometric feature for human identification because of their accuracy and uniqueness. Traditional fingerprint acquisition techniques are contact based and result in poor quality images. The new generation of non-contact based scanners captures high resolution and detailed 3D fingerprint scan, which addresses many of the problems of traditional fingerprint acquisition techniques. The majority of existing fingerprint databases available today are 2D, so there is a need for backward compatibility for the 3D scans captured. In order to solve this interoperability issue, I present an algorithm to unwrap the 3D fingerprint to its 2D equivalent image to be able used in an Automatic Fingerprint Identification System.

Program Committee: Drs. Marc Olano (Advisor, Chair), Penny Rheingans and Dr. Gymama Slaughter

The 2012 Maryland Cyber Challenge (MDC3) will be held this fall with qualification rounds in September and the finals at the Cyber-Maryland Conference in the Baltimore Convention Center on 16-17 October. MDC3 is a statewide cyber competition and conference designed to interest more students and young professionals in pursuing careers in cybersecurity. The goal of the Maryland Cyber Challenge is to help strengthen Maryland’s position as a cybersecurity leader by bringing teams of current and prospective cybersecurity professionals together to develop the skills and techniques needed to protect vital information systems. Founders of the event include SAIC, UMBC, the National Cyber Security Alliance, the Maryland Department of Business and Economic Development, and the Tech Council of Maryland.

UMBC students who are interested in MDC3 or cybersecurity in general are encouraged to contact the UMBC Cyber Defense Team, aka the Cyberdawgs.

MDC3 team registration is now open for the 2012 challenge. Questions about the Cyber Challenge can be sent to Kelli-Ann Tucker (kelli-ann.tucker at saic.com).

MS Thesis defense

Dynamically Reconfigurable Layered Filesystem

Sunil Venkatesh

10:00am Thursday, 26 July 2012, ITE 325b

Traditionally, all files and directories in Linux and UNIX-like systems have been organized in a hierarchical fashion under the root directory “/” adhering to the Filesystem Hierarchy Standard (FHS). Although there is sufficient flexibility in how the filesystem hierarchy is structured given it satisfies the FHS, there is little straightforward means to customize the filesystem structure to suit an individual user’s or a set of users’ needs without affecting rest of the users on a shared system. Our approach aims to eliminate such a restriction by providing isolated environments to individual users with the help of data being organized in the form of layers. Such an environment also provides an important advantage from security perspective by reducing the risk involved in unwarranted access to files by carefully choosing the layers a user has access. Maintainability at the layer level is another key advantage of our approach over the fine-grained approach of dealing with individual files.

Committee: Drs. John Dorband (Chair), Yelena Yesha, Mohamed Younis

MS Thesis Defense

Federating Disjoint Wireless Networks
Using a Mix of Stationary and Mobile Nodes

Ahmad Abbas

12:00PM Thursday 26th July 2012, Room ITE 325b

In many applications need arises to connect a set of disjoint nodes or segments. Examples include repairing a partitioned network topology after failure, federating a set of standalone networks to serve an emerging event, and connecting a sparsely located data sources. Contemporary solutions either deploy stationary relay nodes (RN) to form data paths or employ one or multiple mobile data collectors (MDCs) that pick packets from sources and transport them to destinations. In this thesis we investigate the interconnection problem when the number of available RNs is insufficient for forming a stable topology and a mix of RNs and MDCs is to be used. We present two algorithms for determining where the RNs are to be placed and planning optimized travel routes for the MDCs so that the data delivery latency as well as the MDC motion overhead are minimized. The performance of the algorithm is validated through simulation.

Committee: Professors Mohamed Younis (chair), Ryan Robucci and Tinoosh Mohsenin