CIS Home  |   Penn Engineering  |   Penn

QTM: Quantitative Trust Management MURI




In modern computing, distributed topologies are becoming more prevalent as designers take advantage of the bandwidth, file diversity, and scalability such systems offer. Frequently, nodes have the ability to both request and provide services from other users. This is inherently risky; decentralized models lack the notions of authenticity, reliability, and accountability that monolithic servers can provide. Nonetheless, well-behaved decentralized systems are advantageous.

Behavior recognition and enforcement are the role of trust and reputation systems. The design and study of these programs is the primary focus of the Quantitative Trust Management Multiple University Research Initiative (QTM-MURI). Professors and students at the University of Pennsylvania, Georgia Tech, and Columbia are collaborating to tackle tough QTM problems. Navigate the attached pages to learn more about our work:

  • PERSONNEL - The professors and students behind our work

  • PUBLICATIONS - Our contributions to academic literature.

  • SUMMARY SLIDES - Summary presentations for QTM and each major sub-project.

  • TM/RM SIMULATOR - Simulating a P2P file-exchange system permits the comparative evaluation of trust/reputation management systems. Static network traces are highly configurable, and we include dynamic considerations, such as bandwidth.

  • QuanTM FRAMEWORK - QuanTM, a quantitative trust management system, selectively combines elements from trust management and reputation management to create a novel method for policy evaluation. In particular, one can quantify how much a delegated authorization should be trusted.

  • WIKIPEDIA VANDALISM - By bringing the notions of reputation and feedback to bear on Wikipedia, one is able to detect vandalism without applying natural-language-processing over the article or diff text. Feasability of the technique was shown via an ex-post-facto study -- then built into software which processes Wikipedia edits in real-time.

  • AUTONOMOUS SYSTEM (AS) REPUTATION - AS-CRED is a reputation service for Autonomous Systems (AS), which characterizes their trustworthiness in announcing valid BGP updates. As ASes tend to repeate their behaviors, reputation becomes an effective metric to quantify the level of trust one can place in an AS annoucing a valid BGP update. AS-CRED also uses the computed reputation to identify and trigger alert for any new instance of invalid BGP updates being generated by the ASes. Comparitive studies show that the alerts thus generated from AS-CRED show a dramatic improvement over existing similar alert systems.

  • TRUSTFORGE - The TrustForge project is developing a credentialing scheme and access control subsystem for a crowd-sourcing collaborative environment using he notions of Quantitative Trust Management. The project is funded by DARPA as part of the Adaptive Vehicle Make portfolio of programs.

  • INTERNAL WIKI - (password required) Team member access to internal affairs.

CIS Home  |   Penn Engineering  |   Penn