| Peer to Peer (P2P) networks have become a popular means to connect remote users and share data. Connecting nodes in an overlay network allows them to locate and communicate with one another, search for data, and distribute workload across the network. Privacy has become a significant concern in P2P networks, as users prefer to remain anonymous while sharing or retrieving data. A general approach to achieve anonymity on P2P networks is to construct indirect paths between clients and servers. Popular P2P networks utilizing such an approach include Freenet and Mute. This approach, however, comes at the cost of a large communication overhead, since the most common and resource intensive communication on P2P networks is file transfers.; Significant research has been done towards providing node privacy while maintaining an efficient network. Most approaches relax their anonymity requirements in order to maintain efficiency. Instead of providing full anonymity, protecting the privacy of both clients and servers, these networks provide partial anonymity, providing anonymity to either clients or servers. This allows more lenient implementations to be introduced, which require less network resources. Various implementations exist providing partial anonymity, such as Mantis, Hordes, and Freehaven. In this thesis, we present SHUBAC, a searchable P2P network providing full anonymity. Client and server anonymity is achieved via indirect, tunneled, transfers. A flexible dynamic path approach is introduced that shrinks paths in size to reduce overhead and maintain stability. Paths are dynamically reconfigured throughout a communication to maintain a high level of privacy. This approach enables SHUBAC to make a good tradeoff between anonymity and efficiency. We compare SHUBAC to two other P2P networks. Mantis, which provides partial anonymity but with low overhead, and Mute, which provides full anonymity, but does so with a large overhead. Simulation results show that SHUBAC progressively reduces overhead while maintaining a high level of privacy to both clients and servers. |
