188宝金博页面版

Eric FreudenthalBackground and Research PlanBackgroundI enjoy designing and constructing systems that achieve robustness by adapting to changing constraints. Myrecent focus has been in decentralized systems, including mechanisms for expressing and enforcing securityrelationships among mutually distrustful administrative domains, securely deploying mobile agents, and theefficient dissemination of on-line content. My hope is that this work will facilitate the automated, safe, andefficient deployment of Internet-scale services onto the global base of hitherto autonomous network-connectedsystems. My background also includes computer architecture, interprocess coordination, computer visionand electrical design. Many problems have analogues in other domains, and my relative breadth providesa rich set of alternative frameworks to draw upon when tackling new problems. Below I summarize mycontributions to four research efforts and describe several potential directions for future research.Coordination of Shared-memory SystemsMy dissertation research[3] explored the dynamic interaction of architecture and algorithms. To supporthigh performance busy-wait coordination, the NYU Ultracomputer architecture includes routing switchesthat parallelize concurrent accesses to the same “hot spot” memory addresses through a technique calledhardware combining. My contributions include enhancements to the Ultracomputer combining switch thatsignificantly reduce memory latency and centralized coordination algorithms that, when executed on systemswith combining, have performance competitive with or superior to the commonly used alternatives[4, 5,1]. Attributes of these coordination problems and their solutions have clear applicability to dynamicallyorganized Internet-scale systems.Image RecognitionThe DARPA-sponsored MSTAR effort engaged approximately one hundred scientists at ten institutions inthe construction of an experimental model-based system to detect and identify targets in SAR (syntheticaperture RADAR) imagery. I initially served as technical lead for the NYU effort, and later as the leadinvestigator for two follow-on efforts spanning multiple institutions. My research contributions includedalgorithms for efficient registration and object identification[10], the development of a parallelized hypothesisevaluation and refinement executive[11], and optimizing template selection algorithms[2].Security Infrastructure for Large and Decentralized SystemsI presently lead several security and middleware related efforts in Vijay Karamcheti’s Parallel and DistributedSystems Group (PDSG). These efforts principally investigate the security needs of mobile agents deployedinto dynamic environments that span a large number of administrative domains. Techniques I discoveredto represent and aggregate partial authorization can simplify the administration of large and decentralizedaccess control systems and may be of more general utility.The deployment of and communication between mobile agents requires the establishment of sustainedauthorizing trust relationships between agents and systems that host them, and other agents with whomthey interact. Existing component-based frameworks (e.g. J2EE and grid) do not offer appropriate securityguarantees for coalition systems that span multiple mutually-distrustful administrative domains. In orderto address these challenges, I designed the DisCo deployment substrate and dRBAC, a decentralized role-based access control system. For a DARPA seedling effort in quantified trust management, I am examiningmechanisms for trust aggregation that may increase the expressiveness and scalability of access controlsystems. This work is summarized below.1