Computer Science - Research at Wake Forest University

Jennifer J. Burg

Collaborative Research: Computing in the Arts – A Community-Building Initiative

Awarded $199,750 for the period 9/15/13 to 8/31/16

Source: National Science Foundation (NSF)

Three partner institutions – the College of Charleston, WFU, and UNC Asheville – aim to build a strong, diversified community of computer science educators interested in adopting and developing innovative instructional material related to computing in the arts (cita). The collaboration will disseminate the successes of the cita model implemented at the College of Charleston; synthesize this work with complementary approaches at the other institutions; and create and disseminate innovative instructional materials. Over the three-year grant period, they will host annual faculty development workshops, organize meetings and sessions at the Special Interest Group in Computer Science Education (SIGCSE), and develop a website offering course exemplars and instructional material.

Linking Science, Art, and Practice in Digital Sound

Awarded $396,204 for the period 1/1/08 to 12/31/10

Source: NSF

This collaboration among a computer scientist and an education researcher at a liberal arts university and a digital sound designer at a performing arts conservatory will create, implement, and disseminate innovative curricular material that integrates the science and art of digital sound design. It supplements a traditional textbook with interactive online tutorials, worksheets, MATLAB exercises, programming assignments, and creative and research projects. The curriculum includes mathematical, scientific, and algorithmic explanations of what sampling, quantization, aliasing, dynamic range, imaging, resonance, transforms, filters, normalization, and compression mean in the context of digital audio. Student problem-solving strategies are observed directly and via audio/videotapes. The project’s innovative assessment strategy includes 3 summer workshops for faculty and pilot-testing sites across the United States. The material will be freely available on the web.

CPATH: Revitalizing Computer Science Education through the Science of Digital Media

Awarded $237,991 for the period 8/1/07 to 07/31/10

Source: NSF

This exciting new model of computer science education focuses on the science of digital media, building on natural interdisciplinary relationships, tight links between science and application, and continuous feedback from educators and employers. In recent years, enrollment in postsecondary computer science programs has been slipping. This project looks at computer science education from a new perspective – placing the study of digital media at the center and exploring its relationships to various disciplines and applications. When taught as computer science, digital media can be presented both rigorously and engagingly. The grant will support workshops around the United States, involving representatives from academia, business, and industry, and creating a broad network of stakeholders in computer science education. More coherent and robust models for interdisciplinary programs, particularly those coupling computer science with art, will be developed, and specific ways to link science with practice will be identified. The information gleaned from the workshops will be compiled, analyzed, and reported both incrementally, for continuous refinement, and at the end of the project, for broad dissemination. The interactions among computer scientists, artists, and practitioners in the workshops will generate a new model of computer science education especially appropriate to smaller schools or liberal arts universities, where interdisciplinarity is supported.

Samuel Cho (see also PHYSICS)

Extrapolating the concept of protein corona for understanding nanoparticles at large

Awarded $93,491 for the period 8/15/12 to 7/31/15

Source: NSF/Clemson University

These studies aim to decipher the basic mechanisms of biocorona formation using molecular dynamics (MD) simulations (Cho lab) validated by experiment (Ke lab). Results will translate well-established theoretical and simulation approaches from protein folding to NanoEHS for great educational, economic, and environmental benefits. The Cho lab develops and performs MD simulations to examine the formation of AgNP-biocoronas at molecular resolution and to define the main determinants, such as time, free energy, and stability. Their GPU-optimized approaches perform 10-100x faster than traditional CPU approaches that preclude simulations at the relevant time and length scales. The Ke lab will validate these calculations experimentally; then additional simulations will guide the design of new NP-biocorona simulations. The Cho lab will also model NP binding interactions with proteins and glucose.

Errin Fulp

with Ken Berenhaut, MATHEMATICS

Modeling Mobile Agent Populations and Movement for CEDS

Awarded $196,473 for the period 12/13/10 to 10/31/13

Source: Pacific Northwest National Laboratory (PNNL)/Battelle Memorial Institute

Mathematical models will be used to clarify the basic performance trade-offs associated with agent populations, pheromone strength, and nonregular geographies. As more advanced features are added to the system, simulation will be used to find appropriate settings. Modeling will also be used to characterize and assure the operational security of the Digital Ants system.

An Evolution-Inspired Approach for Moving Target Defenses

Awarded $197,420 for the period 10/1/12 to 9/30/14

Source: NSF

This project aims to develop a prototype Moving Target (MT) defense system, where computer configurations automatically evolve for improved security and diversity. Operating systems and/or applications are modeled as chromosomes, with their settings as individual traits, using genetic algorithms (GA). The proposed environment has three advantages over current MT techniques. First, many MT techniques seek only to hide poorly configured computers, while the proposed system constantly searches for and employs more secure configurations by discovering, emphasizing, and recombining previously known good configurations. Second, GAs rely on mutations to avoid local minima or maxima; here, mutations can encourage diverse configurations across computers and successive configuration implementations. Finally, this approach can be applied to a variety of configurations.

Leap Ahead

Awarded: $54,000 for the period 5/23/11 to 3/8/12

Source: PNNL

Wake Forest University faculty, staff, and students will collaborate with the Pacific Northwest National Laboratory to create a biologically inspired system that allows cooperative adaptation of successful configuration postures across computers while maintaining a desired level of diversity. These postures may be informed by a collaborative analysis system called Vulcan that provides new information on threats and remediation methods. Specifically, WFU will define methods for Digital Ants to extract information from the Vulcan database and transform it to construct new policies and new types of sensors.

Beyond Snort: Extending malicious activity detection with SVMs

Awarded $33,662 for the period 3/1/11 to 8/31/11

Source: Department of Energy (DOE); Lawrence Livermore National Lab

Computer networks are protected by several security measures. For example, an intrusion detection system (IDS) detects malicious behavior internally and externally, and its de facto standard is Snort, which monitors and analyzes network traffic against a user-or community-defined set of signatures, often content-based, that look for malicious behavior. While users can update this signature set, for real-time cybersecurity, it is essentially static. Adversaries try to evade detection by morphing the content and related attack parameters.

Support Vector Machines (SVMs) are supervised learning techniques that show excellent performance in classification tasks. In this study, they are used to classify malicious vs. harmless behavior as defined by a set of Snort signatures. Robust additional features of network sessions found in flow summary data, including control packet sequences, will be used to investigate whether they can re-identify malicious activity that has been modified in an attempt to thwart signature-based detection. System performance will be measured on data sets obtained from Enterprise-level network traffic, using such standard machine learning metrics as accuracy, false positives, and false negatives.

DigitalAnts in Heterogeneous Enclaves

Awarded: $10,000 for the period 2/2/11 to 3/31/11

Source: Battelle Memorial Institute

Wake Forest University faculty, staff, and students will collaborate with Battelle Memorial Institute, Pacific Northwest Division, and University of California, Davis, to develop a security architecture for GENI. Specifically, WFU will design means to accommodate the wide variety of machine and operating system configurations and configuration changes expected in GENI. To accommodate emerging knowledge of zero-day exploits, a mechanism must be in place for the human supervisor to guide DigitalAnts operations via policy updates.

Modeling Mobile Agent Populations and Movement for GENI

Awarded $9,487 for the period 10/1/10 to 12/11/10

Source: Battelle Memorial Institute

Wake Forest University will collaborate with Pacific Northwest National Laboratory (PNNL) to migrate TDMAA code to an EMULAB framework and provide a plan for implementing it; provide a written report on the role of digital ants in the EMULAB-derived deployment; update and deliver the TDMAA code-base to PNNL; develop an EMULAB testbed using WFU’s DEAC-cluster; and implement a small JADE testbed for code development. A WFU graduate student will provide the work to PNNL.

Securing the next generation of information infrastructure

Awarded $22,189 for the period 8/27/09 to 9/30/09

Source: DoE/Battelle Memorial Institute

Dr. Fulp will work with the Pacific Northwest National Laboratory to develop next-generation computing technologies supporting secure command/control and information infrastructures and predictive defense and adaptive systems.

Integrated Parallel Firewall and IDS for High-Speed Networks

Awarded $40,292 for the period 8/1/08 to 8/7/09

Source: DoE; Greatwall Systems, Inc.

This project aims to develop a new, scalable network firewall and Intrusion Protection System (IPS) that can manage increasing traffic loads, higher network speeds, and strict Quality of Service (QoS) requirements. Firewalls remain the frontline defense for securing networks vital to private industry, government agencies, and the military. However, they can easily become bottlenecks; packets must be inspected and compared against complex rule sets, tables, and signatures. As a result, the firewall and, more important, the network it protects are susceptible to Denial of Service (DoS) attacks, which attempt to saturate the firewall with legitimate traffic. This project addresses these crucial security problems using firewall policy optimization and parallelization to provide an affordable, scalable, high-speed firewall and IPS.

Securing the Next Generation of Information Infrastructure

Awarded $35,250 for the period 1/15/08 to 7/30/08

Source: Battelle Memorial Institute, Pacific Northwest Division

Dr. Fulp will work with the Pacific Northwest National Laboratory (PNNL) in research and development of next-generation secure computing technologies supporting secure command/ control infrastructures and information infrastructures, drawing on his expertise in next-generation, high-speed, and quality of service (QoS)-enabled networks and activities based on network pricing and auction research, QoS research, resource-allocation research, and peer-to-peer trust systems, focusing on the security requirements of information and command/control infrastructures and in support of both predictive defense and adaptive systems.

Firewall Architectures for High-Speed Networks

Awarded $51,334 for the period 9/15/05 to 9/14/06

Source: DOE

As network technology advances and becomes more ubiquitous, firewalls must perform important security inspections under increasing traffic loads, faster network connections, and strict quality of service (QoS) requirements that render them susceptible to ottlenecks and denial of service (DoS) attacks. This project investigates a new architecture called hierarchical firewalls in which traffic is quickly distributed among machines based on perceived threat. Traffic considered safe is promptly forwarded into the secure network, while what remains is forwarded to different machines in the hierarchy for further scrutiny. Hence, traffic is segregated and queued based on security threat, yielding minimal delays for legitimate traffic, and the system is robust and highly available, since it uses multiple machines. Preliminary results indicate that it is 6 times faster than any other current firewall system and applicable to a wide variety of agencies in the public and private sectors.

David J. John, with Jacquelyn Fetrow, Computer Science and Physics, and Edward E. Allen, Mathematics

Algebraic and Statistical Models of Redox Signaling

Awarded $123,379 for the period 4/1/08 to 3/31/09

Source: NIH

An interdisciplinary research group spanning the Reynolda and Health Sciences campuses aims to develop theory, algorithms, computational tools, and research methodologies for network modeling of redox-regulated events in human cells. Recent research indicates that redox-regulated networks are central to cellular communication under a variety of normal and diseased conditions, including cancer, neurodegenerative diseases, and aging. This project will (1) identify a comprehensive set of cellular proteins modified at cysteine residues as a result of redox-dependent signaling; (2) correlate the concentration of a given cellular perturbant and its associated redox signal; 3) associate networks with particular perturbants; and 4) produce both topological and dynamic models of the cellular network associated with these pathways. These models will then be compared to other data on protein/protein interactions and kinase cascades to produce a more comprehensive model of cellular regulation and its biological outcomes.

Victor Paúl Pauca

2016 Boeing Analytics Research Scholar Program

Awarded $44,148 for the period 3/1/16 to 8/31/16

Source: Boeing Company

This project aims to develop a tool that can analyze acoustic emission data taken during a structural test and identify the types of failure (fiber breakage, matrix cracking, delamination) occurring throughout the test. It will elucidate the progression of composite failure in order to reduce testing costs and cycle time during the development of new composite material systems and structures. Experts will analyze the data amassed during the High Fidelity Test project and recommend a solution, which will be incorporated into a software tool for use by Boeing structural engineers.

Analytics of Acoustic Emission Data

Awarded $49,645 for the period 1/26/15 to 11/26/15

Source: Boeing Company

The project aims to develop a tool that analyzes acoustic emission data and identifies the types of failure— fiber breakage, matrix cracking, delamination— throughout a structural test. By defining composite failure progression, it will reduce testing costs and cycle time when developing new composite material systems and structures.

with Robert Plemmons, MATHEMATICS/COMPUTER SCIENCE, and Todd Torgersen, COMPUTER SCIENCE

Implicit Geometry and Linear and Nonlinear Tensor-Based Compression and Restructuring of High-Dimensional Multimodality Data Sets

Awarded $61,599 for the period 1/1/13 to 8/25/13

Source: US Department of Defense (DOD)/Boeing Company

The project aims to develop a novel framework based on implicit geometry (IG) and linear and nonlinear tensor decomposition for effective compression and restructuring of multiple-source imagery. This form of compressed representation enables object characterization, target identification, and temporal tracking directly in the compressed domain. IG technology developed at Boeing is extremely effective for flexible 3D data representation with minimal loss of fidelity. Here, IG is used to enable fast compression of LIDAR data, while facilitating object classification and identification.

Analysis of Ultrasound Signal Reconstruction

Awarded $4,230 for the period 10/6/11 to 11/30/11

Source: Boeing Company

WFU will investigate methods for ultrasonic signal deconvolution, including evaluation and/or improvement of Beoing’s compressive sensing algorithm.

Challenging Ocular Image Recognition (COIR)

Awarded $79,477 for the period 4/4/11 to 7/31/11

Source: Intelligence Advanced Research Projects Activity (IARPA)/Carnegie Mellon University

The Wake Forest University team will work alone and in collaboration with teams from West Virginia University, Carnegie Mellon University, and Catholic University of America to achieve (1) robust, multispectral ocular recognition algorithms; (2) fast iris and ocular classification algorithms as part of an effective multiscale recognition approach, drawing on Scale-Invariant Feature Transform (SIFT); (3) effective, efficient algorithms fusing ocular imagery acquired by thermal and visible-light color sensors and image-reconstruction and superresolution algorithms; (4) algorithms to improve recognition of video ocular image sequences; (5) robust segmentation and matching routines for ocular image recognition exploring use of a fast segmentation routine for processing nonideal irises; (6) information-theory methods for estimating the performance of ocular image recognition under other nonideal conditions; (7) software to enable fast, accurate simulation of ocular imaging under nonideal scenarios and first-order performance estimates of the recognition algorithms; (8) a test-bed system to facilitate collection of real face and ocular image data under the same or similar conditions as assumed in simulated image data; and (9) C++ software for each of the above tasks.

A Practical, Enhanced-Resolution, Integrated Optical Digital Imaging Camera

Awarded $14,979 for the period 5/18/09 to 6/30/09

Source: Defense Microelectronics Activity; Catholic University of America

The PERIODIC program aims to significantly advance technology and transitions using computational array imaging. By integrating many optical imaging channels into a single platform with a computational back end, a single high-resolution, high-dynamic-range, extended depth-of-field, polarimetric, and multispectral image is digitally and selectively fused. Simulation work will analyze and develop a new PERIODIC camera, able to produce super-resolved, extended depth-of-field imagery.

with Robert Plemmons

Combining Imaging and Nonimaging Observations for Improved Space Object Identification

Awarded $29,992 for the period 4/1/08 to 11/30/08

Source: Air Force Office of Scientific Research (AFOSR); University of New Mexico subcontract

Present-day imaging and nonimaging capabilities are often inadequate to determine the detailed properties of the ever-smaller satellites increasingly deployed at a variety of altitudes in space. This low-dimensional parametric approach jointly models the essential literal and nonliteral characteristics of space objects in terms of a relatively small set of physically motivated parameters. Digital postprocessing of the data aims to estimate the values of those parameters using polarimetric and spectral data to overcome the raw resolution limits of even the largest existing and foreseeable AF/DoD assets.

Systematic Development of Quantum Computational Software

Awarded $45,000 for the period 2/1/07 to 1/31/08

Source: North Carolina Biotechnology Center (NCBC)/Targacept

Ab initio molecular dynamics (aiMD) is one of the most powerful simulation tools for studying material and biological complex systems. However, novel applications and greater need for efficiency place a tremendous burden on existing aiMD software packages, increasing their size and complexity and making them harder to change and to maintain. QUEST, a novel software system and programming paradigm, will change the way highly complex aiMD algorithms are implemented. With partner company Targacept, Inc., the WFU team will demonstrate how QUEST can greatly reduce the cost of transitioning aiMD technology to biotechnology research and provide a common language to spark further improvements. The work will have a profound impact on drug discovery and molecular recognition and benefit the scientific community at large.

Enhanced Iris Recognition Systems for Personal Identification

Awarded $80,000 for the period 4/1/06 to 9/30/06

Source: DynCorp International, LLC

Computational Methods for Quantum Molecular Dynamics

Awarded $16,000 for the period 10/1/06 to 11/30/07

Source: National Institutes of Standards and Technology (Targacept)

Over the last decades, molecular dynamics and powerful computer technology have been combined to study the dynamic properties of molecules, solids, and liquids. The Car-Parrinello method, a unifying approach for electronic structure calculations based on density functional theory and classical molecular dynamics simulations, enabled more accurate studies of molecular systems without requiring an a priori choice about the nature of the system. Since its appearance in 1985, the Car-Parrinello method remains one of the most influential and widely used for first-principles molecular dynamics. However, its computational requirements are extensive. Even a very short (picosecond) simulation of a small molecule of fewer than 30 atoms requires several weeks of processing. This project aims to develop and to optimize high-performance software for Car-Parrinello molecular dynamics simulations that can substantially increase their length and size for life-science applications and drug design. New approaches that will increase computational and memory efficiency will be explored for both single processor and parallel implementations. In particular, top-down and performance analyses of a base Fortran implementation will seek opportunities for optimization through program transformations and parallelization. Algorithmic transformations that can further reduce the overall computational cost will be studied, focusing on an efficient, modularized implementation.