Eric Freudenthal

Assistant Professor University of Texas at El Paso Computer Science Department Office: Room 202a Lab: Room 320 234 Hawthorne St.(# 36 on the campus map) El Paso, TX 79902 shared fax: 915/747-5030 tethered: 915/747-6954 efreudenthal @ utep.edu wireless: 917/279-6208

• 10:30 AM on Fridays
• 7:30 PM on Tuesdays
• Let me know of your intention to visit (send email) since I will depart if students are neither present nor expected.
• Please send email suggesting alternate times if this schedule is not convenient.

Scheduling a meeting: Since my public calendar is available online, you can select and propose a mutually convenient meeting via email. It's particularly convenient if this invitation is generated by a calendar tool.

• For Outlook: send invitations to efreudenthal@utep.edu
• For Google Calender: send invitations to eric@freudenthal.net

I encourage your invitations to include (1) your name, (2) preferred email address, (3) phone number and (4) reason for the meeting. It's also a good idea to suggest alternate times.

Research

I enjoy designing systems that achieve robustness through adaptation to changing constraints. My recent focus has been in decentralized systems, including mechanisms for expressing and enforcing security relationships among mutually distrustful administrative domains, securely deploying mobile agents, and the efficient dissemination of on-line content. My background also includes electrical design, architecture, interprocess coordination, and computer vision.

A summary of my research appears below. A more formal research statement written in Spring 2004 (in Acrobat format) is also available.

New Efforts (updated Summer 07)

• Toothless - Ryan Spring (A PhD candidate) is investigating techniques useful for constructing a cooperative read-only filesystem that we anticipate will provide better performance for common workloads than local disks. Toothless was motivated by our desire to minimize the administrative overhead to provide high performance services on ReCoN (see below). Like Google's GFS, Toothless will achieve high performance through the tasteful selection of relevant filesystem semantics. No homepage yet, but Ryan's working on it.
  
• Modular Network Protocol Stacks
  In collaboration with Nancy Griffeth at Lehman/CUNY, I am investigating techniques to enable the implementation, performance evaluation, and formal verification of a netowrk protocol from a single specification.
• Addressing the challenge of teaching systems skills to students who learn Computer Science in Java (curriculum reform) Students who learn CS in Java are frequently not familiar with memory management, pointer manipulation, and runtime support, and thus are under-prepared for serious systems development work. This reformed 4^th semester course simultaneously teaches C and assembly language and thus also exposes the students to runtime support for high level languages and prepares them to develop device drivers, operating systems, and embedded systems. (homepage coming soon)

Ongoing Efforts

• Fern - An updatable authenticated dictionary for autonomic systems
  Authenticated dictionaries are useful for disseminating authorization and other security-relevant documents in distributed environments. Fern provides the strong integrity properties of others' authenticated dictionaries in a manner that permits efficient dissemionation of updates upon autonomic content distribution networks such as CoralCDN. (read more)
• ReCoN : A Reconfigurable Computing & Network Lab
  This CRI-funded lab utilizes virtualization to permit students to safely experiment with system configuraton and non-trivial network configuration. (both a research effort and curriculum development; homepage coming soon)

  Prior Efforts

  
  I collaborated with Michael Freedman and David Mazieres in the development of CoralCDN, a locality-sensitive self-organizing content dissemination network. Coral's indices are stored in a hierarchy of interleaved distributed hash tables that share the same name space. Constituent hash tables represent nested ranges of network locality constraints, and all nodes are members of a global hash table with no locality constraints. A single Coral node represents the same hash bucket in multiple hash tables, and searches prefer to search tables with better network connectivity, and only revert to tables with inferior connectivity when necessary.

  Echoing characteristics of the Ultracomputer's combining network, Coral dynamically replicates data near to clients, thereby minimizing hot-spot congestion. While Coral is not robust to security challenges, it is expected to to provide high performance even in the presence of partial system failure.

  More information on this project is available on the Coral home page.

• DiSPAC -- Making the Internet Safe for Grandma
  I have been investigating techniques that can reduce the extreme vulnerability of systems maintained by naive users to the ever-growing threat of malware. This project, tentatively named DiSPAC (Distributed Security Permissions for Alien Code), provides mechanisms that enable naive users to define sandbox execution policies that rely on trusted third parties (functioning as utilities) to determine program characteristics. A web site describing this effort is under construction: http://rlab.cs.utep.edu/dispac.
• Security Infrastructure for Decentralized Systems
  I recently led an effort of the NYU Parallel and Distributed Systems Group (PDSG) to investigate the security needs of systems deployed into dynamic environments that span a large number of administrative domains.

  The deployment of and communication among dynamically deployed software agents requires the establishment of sustained authorizing trust relationships between agents and systems that host them, and other agents with whom they interact. Existing component-based frameworks (e.g. J2EE and grid) do not offer appropriate security guarantees for coalition systems that span multiple mutually-distrustful administrative domains. In order to address these challenges, we developed a deployment substrate for mobile agents called DisCo and a decentralized role-based access control system called dRBAC. I am also investigating quantified trust management, that includes mechanisms for trust aggregation that may increase the expressiveness and scalability of access control systems.

  An extended summary of this work is available online at http://rlab.cs.utep.edu/~freudent/pdsg.html.
  
  

• Coordination for Shared Memory Systems
  As a graduate student supervised by Allan Gottlieb, I investigated support for scalable inter-process coordination on shared-memory MIMD systems. My contributions include detection and analysis of problems in architectures that implmement hardware combining. I propose design modifications that significantly mitigate these effects. I also have contributed centralized algorithms that have lower synchronization latency than those previously known (and superior to commonly used alternatives).

  Hot spot contention in combining networks investigated in my research has analogues in other networked systems. I anticipate that variants of the techniques I propose to mitigate the impact of hot spot congestion on both hot spot and non hot spot traffic can be generalized to other networked systems.

  A more complete summary of my dissertation reseaerch is available online: http://rlab.cs.utep.edu/~freudent/thesisSummary.html. Additional details are available in Technical Report TR2003-849. This report and my full dissertation can be downloaded from the NYU Computer Science Department web site.
  
  

• Image Recognition
  I investigated automatic target recognition in imagery collected using synthetic aperture radar, participating in several research projects associated with DARPA's MSTAR model-based vision research program and the AFRL's Model Based Vision Lab. I collaborated with Lockheed-Martin, Veridian-Erim, Diamondback Vision, and the University of Cincinatti on the MEP4 project that investigates identification of partially occluded targets. I also collaborated with Alphatech Corporation and SAIC to investigate the inherent complexity of the SAR ATR problem. This project was awarded a second phase STTR.

  In collaboration with Ben Goldberg and Davi Geiger, I organized the NYU Recognition Lab, computational resource available for research in computer vision as applied to automatic target recognition. The equipment for this lab was purchased under a grant from the AFOSR's DURIP program.

  The DARPA-sponsored MSTAR effort engaged approximately one hundred scientists at ten institutions in the construction of an experimental model-based system to detect and identify targets in SAR (synthetic aperture RADAR) imagery. My research contributions included algorithms for efficient registration and object identification, the development of a parallelized hypothesis evaluation and refinement executive, and optimizing template selection algorithms.