@MASTERSTHESIS\{IMM2004-03254,
    author       = "T. Cholewinski",
    title        = "Probalistic public-key confidence valuation model for a peer to peer {PKI}",
    year         = "2004",
    keywords     = "Peer to Peer, P2P, Public Key Infrastructure, {PKI,} {PGP} Web of Trust, Trust, Authenticity, Java",
    school       = "Informatics and Mathematical Modelling, Technical University of Denmark, {DTU}",
    address      = "Richard Petersens Plads, Building 321, {DK-}2800 Kgs. Lyngby",
    type         = "",
    note         = "Supervised by Assoc. Prof. Christian D. Jensen",
    url          = "http://www2.compute.dtu.dk/pubdb/pubs/3254-full.html",
    abstract     = "Peer to Peer systems are becoming widespread throughout the Internet and pervasive computing systems. Existing {PKI} infrastructures - both hierarchical and non-hierarchical cannot be directly ported to P2P environments. This is because existing PKIs rely heavily on the presence of CAs, which act as a {''}trusted third party{''} in the system. The problem of feasibility of implementation of various functions performed by {PKI} systems in such an environment is analyzed.

The goal of this thesis is to explore the possibility of implementing a Peer to Peer {PKI} system based on the idea behind the {PGP} Web of Trust and a probabilistic algorithm to evaluate the con dence in a public-key from a {CA,} which is described in the paper {''}Modeling a {PKI} Infrastructure{''} by Ueli Maurer [19]. The public-key valuation model makes assumptions about trust in the {''}trusted third party{''} explicit, which allows the system to use key-servers that are not completely trusted. This is particularly helpful in a wireless P2P environment targeted by this work.

The feasibility of implementation of the probabilistic confidence parameter valuation model is evaluated using a software prototype.

The conclusions drawn during the design and implementation phases of the prototype serve as a basis of an overall feasibility evaluation. A problem is identified involving the complexity of calculations of higher level trust paths. Further research paths are outlined, including sensitivity analysis for  nding certi cation paths which contribute most to the end-value of the confidence parameter."
}