From dralansherman@starpower.net Thu May 1 01:23:48 2008 Date: Thu, 1 May 2008 00:33:16 -0400 From: Dr. Alan T. Sherman To: CSEE ALL Subject: [Csee-faculty-lecturer] CSEE Research Review - Talk Abstracts (revised) CSEE Research Review - Talk Abstracts Friday, May 2, 2008 Department of Computer Science and Electrical Engineering University of Maryland, Baltimore County (UMBC) Session I Penny Rheingans, VANGOGH Lab Scientific Volume and Flow Illustration Data visualization is the creation of visual representations of large data sets in order to facilitate exploration and understanding. Accurately and automatically conveying the structure of a volume model is a problem not fully solved by existing volume rendering approaches. Standard volume rendering approaches create images that may match the appearance of translucent materials in nature, but may not embody important structural details. We have introduced the volume illustration approach, combining the familiarity of a physics-based illumination model with the ability to enhance important features using illustration-inspired rendering techniques. Since features to be enhanced are defined on the basis of higher-order model characteristics rather than volume sample value, the application of volume illustration techniques requires less manual tuning than the design of a good transfer function. Volume illustration provides a flexible unified framework for enhancing structural perception of volume models through the amplification of features, the addition of illumination effects, and the application of procedural textures. These methods can be extended to capture the movements of features in time-varying volumes. I show examples from medicine, computational fluid dynamics, and atmospheric physics. Dhananjay S. Phatak, Cyber Defense Lab Spread-Identity Internet Architecture The Spread-Identity (SI) Internet Architecture is illustrated. It is shown to simultaneously mitigate/solve multiple problems including: (1) Distributed Denial of Service (DDOS) resilience (multi layer multi pronged DDOS defenses as well as offenses) (2) Extremely fast misbehavior detection simply by IP level address (aka token) matching. (3) Drastically mitigating address scarcity in a manner transparent to both the end users as well as the core infrastructure. (4) Substantially simplify network traceback. More generally substantially simplify the control plane and collaborative filtering (since destination address can itself be leveraged as a flow marker). (5) Improve the edge-to-edge traceback while simultaneously bolstering the end-to-end anonymity. (6) Improve the overall network security. A deeper look indicates that spread identity principles are more fundamental and pervade most systems, human-made as well as natural. Konstantinos Kalpakis Energy Efficient Data Gathering and Aggregation in Wireless Sensor Networks Wireless adhoc networks of micro-sensors (WSNs) are proliferating rapidly and transforming how information is gathered, and processed, and acted upon. Sensors are typically inexpensive devices consisting of sensing, data processing, and communication components. Sensors typically operate in unattended mode, communicate with one another over short distances, and establish multi--hop communication routes to one or more base stations. The limited energy, the large number of sensor nodes, the hostile working environments, and the nature of unpredictable deployment of wireless sensor networks have introduced many interesting challenges. Motivated by continuous monitoring applications of wireless sensor networks with non--replenishable energy, we considering the problem of continuous gathering and in-network aggregation of sensor measurement data to the base station with the goal of maximizing the system lifetime. The system lifetime is the earliest time any sensor depletes its energy. We present efficient and effective combinatorial algorithms for the data gathering and in--network aggregation problem in WSNs. Time permitting, we present small-space and easily-computable summaries of sensor data, which are projections on few top eigenvectors of certain query matrices, and their application in computing approximate answers to value-range and location-range queries. Session II Yun Peng Probabilistic Reasoning with Uncertain Data Iterative Proportional Fitting Procedure (IPFP) has been proposed to update a high dimensional joint probability distribution (JPD) with a set of constraints in the form of low dimensional distributions. In contrast to data typically used in Bayesian reasoning and learning, these constraints are data of uncertainty. In the recent years, we have developed a suite of algorithms to address several key limitations of the original IPFP. These include 1) extending IPFP to Bayesian networks (BN) to support belief update and BN learning with uncertain data; and 2) modifying IPFP to handling inconsistent constraints for general JPD and for BN. Mohamed Younis Base-Station Relocation for Increased Dependability of Wireless Sensor Networks Wireless sensor networks (WSN) are composed of a number of sensors probing their surroundings and disseminating the collected data to a base-station for processing. WSN have attracted lots of attention in recent years due to their potential in many applications such as border protection and combat field surveillance. Given the criticality of such applications, maintaining a dependable operation of the network is a fundamental objective. However, the resource-constrained nature of sensor nodes and the ad-hoc formation of the network, coupled with often an unattended deployment, pose non-conventional challenges and motivate the need for special techniques for the design and management of WSN. In this talk, we highlight the potential of repositioning the base-station, which acts as a sink node for the collected data, as a viable means for increasing the dependability of WSN. We show that base-station relocation can be very effective in optimizing the network functional and non-functional performance objectives and in coping with dynamic changes in the environment and available network resources. We further discuss a number of interesting research problem that we are currently investigating. John Kloetzli (Advisor: Marc Olano), VANGOGH Lab - Award for Best Research by an MS Student High-Quality Magnification Texture Filtering We present a high-quality texture magnification filtering technique based on piecewise, continuous Bezier triangulations. We approximate different reconstruction filters as regular, shift-invariant Bezier triangulations. By convolving these approximations with textures we construct a resolution-independent preconvolved texture format which can be rendered in real-time. We present two specific triangulations which allow us to reconstruct each piecewise polynomial from a small initial set of weights. Rendering is performed very quickly in a pixel shader and requires only a small number of local, cache-coherent, independent texture accesses. Our method is suitable for hardware support because it is general, fast, requires only two to four times the space of a standard texture, and works with current texture minification methods. Session III Hillol Kargupta, DIADIC Research Lab Changing the World: Breakthroughs in Communication and the Next Generation of Data Mining This talk will revisit some of the breakthroughs in the field of communication that changed the human civilization and try to understand where we are going tomorrow. It will argue that while we have become very good in quickly connecting an entity with another entity world-wide as long as the former knows the address of the latter, current technology for taking the message or service from one individual to a large population of willing and interested individuals is still very primitive. We need technology for dealing with the new world of attention-economics. The talk will also argue that the currently popular centralized client-server model for the Internet applications may not work well in doing so. It will offer some alternate thoughts borrowed from the nature and some emerging scalable applications that may offer some directions. It will particularly consider some possibilities for peer-to-peer environments and discuss a collection of challenges for the next generation of data miners. Russell Fink (Advisor: Alan Sherman), Cyber Defense Lab - Award for Best Research by a PhD Student (Joint work with Alan Sherman and Richard Carback) TPM meets DRE: Trusted Platform Modules for Electronic Voting Systems We created a design that reduces the required trusted computing base for Direct Recording Electronic (DRE) voting machines incorporating Trusted Platform Modules (TPMs). Using trusted TPM hardware and some trusted BIOS firmware, our approach cryptographically binds each voter's choices with the presented ballot and the state of the DRE software. At the end of election day, the DRE produces an official record comprising a randomly ordered list of the votes cast, hashed and digitally signed by the TPM, in a way that anyone can verify that the record is unmodified and was endorsed by the TPM during the official voting period. Our technique carefully manages a Platform Vote Ballot binding key within the TPM so that it can be used only during the voting period. Our method reveals unauthorized software, ballot modifications, vote tampering, and creation of fake election records before or after the official voting period. This talk describes the high-level protocol and discusses plans for implementing it using current TPM technology. Fow-Sen Choa, Photonics Technology Lab Near- and Mid-IR Sensing Using Photon Counting Arrays and Quantum Cascade Lasers The talk will cover research works done at Prof. Fow-Sen Choa's group on the subjects of Near IR sensing using avalanche photodiode based photon counting arrays and Mid-R sensing using quantum cascade laser based photo-acoustic sensing. [ Part 2: "Attached Text" ] _______________________________________________ Csee-faculty-lecturer mailing list Csee-faculty-lecturer@cs.umbc.edu http://www.cs.umbc.edu/mailman/listinfo/csee-faculty-lecturer