Department of Computer Science and Electrical Engineering

CRA-W Distinguished Lecture Series
University of Maryland, Baltimore County (UMBC)

Computing for Good: Experiences in Humanitarian Computing

Ellen Zegura
Professor and Chair School of Computer Science
College of Computing, Georgia Tech

4:00 – 5:00 PM, Monday, April 16, 2012, Albin O. Kuhn Library Gallery

Almost four years ago, I was involved in the creation of Computing for Good (C4G), a Georgia Tech College of Computing initiative centered around using computing to help solve pressing societal problems. The primary activity of C4G has been a project-based course taught once per year and taken by seniors (satisfying the capstone requirement) and masters-level graduate students.  Projects with life beyond one semester are frequently taken up by master's students as MS projects.

Over the last year, I have had the opportunity to work closely with the Carter Center and their Mental Health Program in Liberia. With students in the C4G fall 2010 course, my experiences have included technology consulting, technology training, technology integration, and technology invention. I have also had the chance to observe first hand a set of additional country-wide challenges where computing might play a role.  In this talk, I will describe my experiences and highlight additional opportunities for computationalists.

SHORT BIO: Professor Ellen W. Zegura received the BS degree in Computer Science, the BS degree in Electrical Engineering, the MS degree in Computer Science and the DSc degree in Computer Science from Washington University, St. Louis. Since 1993, she has been on the faculty in the College of Computing at Georgia Tech. She currently serves as Professor and Chair of Computer Science. She received an NSF CAREER Award in 1995, a Washington University distinguished Alumni Award in 2008, and was selected as an IEEE Fellow in 2010. She was elected to the CRA Board of Directors in 2011.

Professor Zegura has conducted research and taught in computer networking for over 20 years. Her research interests include the Internet, with a focus on its topological structure and services, as well as mobile wireless networking. In network topology, she is the co-creator of the GT-ITM suite of Internet topology modeling tools, which remains in use 15 years after its original release. In mobile wireless networking, she and her colleagues invented the concept of message ferries to facilitate communications in environments where network connectivity is unreliable and/or sparse. Almost four years ago, she helped create the Computing for Good initiative in the College of Computing, a project-based teaching and research activity that focuses on the use of computing to solve pressing societal problems.

Host: Professor Marie desJardins

EE Graduate Seminar

Unsupervised Multispectral Image Classification

Shih-Yu Chen
PhD (EE) Student, CSEE Dept/UMBC

11:30am-12:45pm, Friday, 6 April 2012, ITE 237

This seminar presents a new approach to unsupervised classification for multispectral imagery. It first uses a Gaussian pyramid multi- resolution technique to reduce image size from which the pixel purity index (PPI) is implemented to find regions of interest (ROIs) with PPI counts greater than zero. These PPI-found samples are further used as support vectors for a support vector machine (SVM) to classify data. The resulting SVM-classified data samples are further processed by a newly designed iterative Fishers linear discriminant analysis (IFLDA), which implements FLDA in an iterative manner to refine classification results. The experimental results show the proposed approach has great promise in unsupervised classification.

Shih-Yu Chen received the BSEE degree from Da-Yeh University in 2005, and the MSEE degree from National Chung Hsing University in 2010. He is currently a PhD (EE) student at UMBC. Mr. Chen's research interest includes medical image, remote sensing image, and vital-sign signal processing.

Host: Prof. Joel M. Morris

Suffix trees for very large strings

Dr. Marina Barsky
University of Illinois at Urbana-Champaign

1:00pm Friday, 30 March 2012, ITE 325b, UMBC

The seminar is dedicated to the construction of suffix trees in external memory. A suffix tree is a compact index of all substrings of a given text. While being asymptotically linear in the size of the input, in practice, suffix trees can easily be 50 times larger than the input. As such, suffix trees often exceed typical main memory sizes, even when the input does not. As most existing algorithms are designed for RAM, their performance severely degrades when the tree and/or input do not fit in main memory. So far, this has prevented the wide application of suffix trees for the analysis of massive string collections.

We will look at new advanced methods of suffix tree construction which circumvent memory concerns and allow us to construct suffix trees for inputs of any size using secondary storage (magnetic disks). We will also discuss how this disk-based index can be used for facilitating the pattern discovery in sequential data.

Dr. Marina Barsky is currently a Post-Doctoral fellow in the Department of Computer Science at the University of Illinois at Urbana-Champaign, IL. She received her PhD in Computer Science from the University of Victoria, British Columbia, Canada in 2010. A large part of her post-graduate research was dedicated to pattern discovery in string data, primarily in massive DNA databases. She is currently expanding her expertise in database systems to new areas such as index management and data mining. Her research interests include data mining of sequential data, information networks, and teaching of computer science through interactive interfaces.

Host: Richard Chang

EE Graduate Seminar

High Power Mid-Infrared Quantum Cascade Laser
Arrays for Standoff Photoacoustic Chemical Detection

Xing Chen, PhD (EE) Student
Computer Science and Electrical Engineering, UMBC

11:30am-12:45pm, Friday, 30 March 2012, ITE 237

Quantum cascade lasers (QCLs) are compact, powerful, mid-infrared, Semiconductor laser sources. High power QCLs are very important to infrared counter measures (IRCM) and standoff chemical detection applications, as well as others. The performance of such systems critically depend on the amount of power that QCLs can produce. One way to achieve high power operation is to use multi-emitter phase-locked laser arrays.

The first part of the seminar presents the issues and challenges to design, fabricate, and characterize multi-emitter phase-locked QCL arrays for achieving high power operation. The second part of the seminar discusses using high power mid-infrared QCLs to perform standoff photoacoustic (PA) chemical detection. The PA effect is a photo-matter effect involving generation and detection of an acoustic signal when a gas sample absorbs electromagnetic energy (particularly of light).

In recent years, with the help of the development of mid-infrared QCLs, significant progress has been made in their use for PA chemical detection, and sensitivity has been improved significantly. Our theoretical and experimental studies of standoff photoacoustic chemical detection, using QCLs as the laser source, will be presented.

Xing Chen received the BS degree in Opto-Electronics Engineering from Huazhong University of Science and Technology in 2007, and the MSEE degree from UMBC in 2009. He is currently a PhD (EE) candidate at UMBC. Mr. Chen's research interest includes design and fabrication of high power mid-infrared phase-locked QCL arrays and application to standoff photoacoustic chemical detection.

Host: Prof. Joel M. Morris

Changing the Landscape of Voting and Voter Registration through Universal Design

Dr. Juan E. Gilbert
School of Computing
Clemson University

12:00-1:00pm Wednesday, 28 March 2012
room 459 ITE Building, UMBC

Subsequent to the debacle of the 2000 U.S. Presidential election, it became abundantly clear that America’s archaic voting system was in dire need of a major overhaul. Consequently, Direct Recording Electronic (DRE) voting machines were purchased by several states. The use of these machines has not been without controversy with respect to security, trust and ease of use. Professors and security research teams have found several vulnerabilities in current voting technologies. In 2002, the Help America Vote Act (HAVA) was created to provide all citizens equal access to participate in the electoral process, regardless of ability. The Prime III voting system, http://www.PrimeVotingSystem.com , is a secure, multimodal electronic voting system that takes a universal design approach to address security, trust and ease of use. Dr. Gilbert and his research team were recently awarded a $4.5 million dollar grant from the U.S. Election Assistance Commission to conduct research on accessible voting technologies.

Dr. Juan E. Gilbert is an IDEaS Professor and Chair of the Human-Centered Computing Division in the School of Computing at Clemson University where he leads the HCC Lab. He is also a Professor in the Automotive Engineering Department at Clemson University. Dr. Gilbert is a Fellow of the American Association for the Advancement Science (AAAS), an ACM Distinguished Scientist, National Associate of the National Research Council of the National Academies, an ACM Distinguished Speaker and a Senior Member of the IEEE Computer Society. In 2011, Dr. Gilbert was given a Presidential Award for Excellence in Science, Engineering and Mathematics Mentoring by President Barack Obama.

Interaction with Virtual Environments

Tabitha C. Peck
Event Lab, University of Barcelona

1:00pm Tuesday 27 March 2012, ITE32bb, UMBC

Immersive virtual environments (VEs) enable user-controlled interactions within a computer-generated virtual world, such as head-controlled point-of-view, user-controlled locomotion, and user-controlled self-avatars. In this talk I will present three projects focusing on the development of VE systems through understanding human interactions within the VE. The first project presents a VE system that enables users to really walk through VEs that are larger than the tracker-space by manipulating the imprecisions of the human visual system. The remaining two projects focus on virtual embodiment. The theory of embodiment is based on the plasticity of the human mind and its ability to accept a virtual avatar’s body as its own. One theory as to why embodiment works, following the same underlying principles thought to cause the “rubber hand illusion” from cognitive psychology, is that when given appropriate visual and/or haptic stimuli, people will accept an external representation of a body part as their own. This effect has been shown to extend to full-body avatars in virtual environments. I will present one project that demonstrates, through electroencephalography (EEG), that people respond to a virtual avatar as if it is their own body, and a second project that explores harnessing the powers of embodiment to reduce racism and study other psychological issues.

My name is Tabitha C. Peck and I am a post-doctoral researcher at the Event Lab in Barcelona, Spain working with Professor Mel Slater. I received my PhD from The University of North Carolina at Chapel Hill under the supervision of Professors Henry Fuchs and Mary C. Whitton. My PhD research focused on locomotion interfaces in virtual environments and enabling people to physically walk in small spaces while walking in much larger virtual spaces. I am currently working in the European project, Virtual Embodiment and Robotic Re-Embodiment (VERE), and my current research focuses on the psychological effects of embodiment in virtual environments. My research interests include immersive virtual environments, virtual embodiment, human-computer interaction, 3D user interfaces, locomotion, navigation, system design and evaluation, and human perception.

Securing Cyber-Physical Systems

Alvaro Cardenas
Fujitsu Laboratories of America

1:00pm Monday 26 March 2012, ITE 325b, UMBC

Our critical infrastructure systems are being modernized with information and communication technologies to face the operational requirements and efficiency challenges of the 21st century. The smart grid in particular, will introduce millions of new intelligent components to the electric grid, buildings, and homes within the next decade. While this modernization will bring many operational benefits to infrastructure systems, it will also introduce new vulnerabilities, a larger attack surface, and raise privacy concerns.

This presentation will be divided in three parts. The first part of the talk will cover the unique and fundamentally new challenges and solutions required for securing cyber-physical systems. The second part of the talk will focus on new mechanisms for securing cyber-physical systems. The final part of the talk will cover my other research interests in intrusion detection and future plans for big-data security.

Alvaro A. Cárdenas is a research staff engineer at Fujitsu Laboratories of America. Prior to this he was a postdoctoral fellow at the University of California, Berkeley working in securing critical infrastructure systems. His research focuses on network security, the smart grid and other cyber-physical systems, intrusion detection and big data security. He has received numerous awards for his research including a best paper award from the U.S. Army Research Office, a best presentation award from the IEEE, a fellowship from the University of Maryland, and a Distinguished Assistantship from the Institute of Systems Research. He has also been an invited visiting professor at the University of Cagliari. Alvaro holds M.S. and Ph.D. degrees from the University of Maryland, College Park, and a B.S. from Universidad de los Andes.

See http://www.csee.umbc.edu/talks for more information

Addressing failures in wireless sensor networks

Krasimira Kapitanova
University of Virginia

1:00pm Wednesday 28 March, 2012, ITE 325b, UMBC

Wireless sensor networks are now being used for a growing number of applications, from mission critical applications, including fire-fighting, emergency response, infrastructure monitoring, and medical application, to smart home applications, such as home automation, energy efficiency, and home security. These applications must operate reliably and continuously due to the high costs associated with system failure and maintenance. However, continuous and reliable operation of sensor networks is notoriously difficult to guarantee due to hardware degradation and environmental changes, which can cause operating conditions that were impossible for the original system designers to foresee. Recent studies have found that low-cost sensors suffer from many types of faults. Inexpensive nodes can break and battery-powered nodes lose power. Furthermore, sensor network installations suffer from a large number of non-fail-stop faults in which the sensor does not completely fail. Instead, it continues to report values, but the meaning of the values changes or becomes invalid. This talk will discuss a number of new run-time techniques that use application-level semantics to detect, assess, and adapt to sensor node failures.

Krasimira Kapitanova is a PhD candidate of Computer Science at the University of Virginia. Her research focuses on wireless sensor network, in particular using formal approaches for event description and detection. She is also interested in how testing and machine learning techniques can be used to improve the reliability of sensor network applications.

Host: Tim Finin

EE Graduate Seminar

From Backpack to Briefcase: Transition from
the Academic World to Corporate Culture

George W. Reynolds
Director, Industry and University Initiatives
Northrop Grumman Electronic Systems

11:30-12:45pm Friday 16 March 2012, ITE 237

This seminar explores the pitfalls that new hires can encounter as they transition to the corporate world from academia. We explore the personal characteristics that people will evaluate, over and above performance. We will learn the importance of having appropriate, unbiased mentors and how to choose them, as well as strategies to develop key relationships that add value to our career growth and non-technical education. The importance of image and appropriate wardrobe for success will be addressed in addition to what top executives look for when selecting staff with leadership potential.

George Reynolds is a licensed professional engineer with over forty years of experience with the Westinghouse Electric and Northrop Grumman Corporations. His current responsibilities include establishing key strategic relationships with selected universities for long-term research, business and recruitment partnerships. He is also responsible for sector wide initiatives that include knowledge management and introducing lean thinking into engineering and manufacturing organizations.

Mr. Reynolds has served as industry liaison for the Lean Aerospace/ Advancement initiatives at MIT since its inception in 1992. He also serves as the chairman and/or member of numerous engineering Advisory Boards for major universities, and is the past Chairman of the Aerospace Industries Associations Engineering Management Committee. Mr. Reynolds was selected as National Black Engineer of the Year for Professional Achievement in Industry in 1991, and Black Engineer of the Year for Corporate Support of Engineering Education in 2008. He is one of three people in the nation to receive two of these awards. In 2001, he was awarded the Distinguished Black Marylander Award.

Mr. Reynolds holds a B.S. in Engineering from Howard University, an M.S. in Engineering Administration from George Washington University, and is a graduate of the Program for Management Development at Harvard School of Business. He is also a Johns Hopkins Fellow in the Management of Change. He holds a black belt in Six Sigma and is an expert in Lean Thinking. Mr. Reynolds holds a commercial pilots license with multi-engine and jet ratings.

Host: Prof. Joel M. Morris

Analytics for Detecting Web and Social Media Abuse

Dr. Justin Ma, UC Berkeley

1:00pm Friday 16 March 2012, ITE 325, UMBC

The Web and online social media provide invaluable communication services to a global Internet user base. The tremendous success of these services, however, has also created valuable opportunities for criminals and other miscreants to abuse them for their own gain. As a result, it is both an important yet challenging problem to detect, monitor, and curtail this abuse. However, the large scale and diversity of these services, combined with the tactics used by attackers, make it difficult to discern one clear and robust signal for detecting abuse. One approach, relying on domain expertise, is to construct a small set of well-crafted heuristics, but such heuristics tend to rapidly become obsolete. In this talk, I will describe more robust approaches based on machine learning, statistical modeling, and large-scale analytics of large data sets.

First I will describe online learning approaches for detecting malicious Web sites (those involved in criminal scams) using lexical and host-based features of the associated URLs. This application is particularly appropriate for online algorithms as the size of the training data is larger than can be efficiently processed in batch and because the features that typify malicious URLs evolve continuously. Motivated by this application, we built a real-time system to gather URL features and analyze them against a source of labeled URLs from a large Web mail provider. Our system adapts in an online fashion to the evolving characteristics of malicious URLs, achieving daily classification accuracies up to 99% over a balanced data set.

Next I will describe our ongoing efforts for creating analytics for detecting social media abuse. Deciding on a universal definition of social media abuse is difficult, as abuse is often in the eye of the beholder. In light of this challenge, we explore a more formal definition based on information theory. In particular, we hypothesize that messages with low information content are likely to be abusive. From this, we develop a measure of content complexity to identify abusive users that shows promise in our early evaluations.

In addition to our own experiments in the lab, this work has found success in practice as well. Companies serving hundreds of millions of users have adopted these ideas to improve abuse detection within their own services.

Justin Ma is a postdoc in the UC Berkeley AMPLab. His primary research is in systems security, and his other interests include applications of machine learning to systems problems, systems for large-scale machine learning, and the impact of energy availability on computing. He received B.S. degrees in Computer Science and Mathematics from the University of Maryland in 2004, and he received his Ph.D. in Computer Science from UC San Diego in 2010.

Host: Anupam Joshi
See http://www.csee.umbc.edu/talks for more information