Distributed Anonymous Information Retrieval: Freenet

We plan to investigate distributed anonymous retrieval systems by
analyzing the Freenet implementation.  Freenet divides up files among
hosts; encrypts them; and replicates them based on their popularity.
Our chief interest is how one can maintain anonymity yet search
engines can still determine where parts of files reside.  How can a
user's machine send data to a machine requesting it without explicitly
resolving it's IP address and forwarding packets with its signature.
We also plan to investigate and perhaps come up with some new ways to
determine when files get deleted and replicated.  The fear is that the
system can become a virtual /tmp directory where many people just dump
data for the convenience.  Any system should be able to quickly clean
these types of files out, yet keep important files around.  Freenet is
the ultimate heterogeneous system.  It is important for a system such
as Freenet to take the node's capabilities and internet connection rate
into account.

We will first get the latest CVS snapshot of Freenet and attempt to
build our own Freenet system at home.  This should take under two
weeks for us to get the main parts of the system performing as
expected.  There are a number of papers published on Freenet which are
freely accessible from the sourceforge website.  We will read these in
order to better understand the architecture of the system, and perhaps
some of its weaknesses.  We expect to encounter a number of obstacles
that will need to be overcome because of the "in development" status
of the system, as many of its major features have not yet been
implemented. 


Key Points
----------

o Scalability

  How well does a system such as Freenet scale with a growing number
  of hosts? Can the increased message passing between hosts become a
  serious bottleneck, such as in other fully distributed systems such
  as Gnutella?

o Anonymity

  Are Freenet objects truly anonymous? With emerging technologies for
  IP traceback (IPv6 for example) how do we keep objects from having
  their origins traced?

o Persistence / Durability

  How does Freenet insert and distribute objects? How does it purge
  objects? What if an object needs immediate removal, such as a trojan
  virus executable?

o Availability / Caching

  How do more popular objects become more accessible (available)?

o Fault Tolerance

  What happens in the case of a node failure, be it temporary or
  permanent?