Freenet: A Distributed Anonymous Information Storage and Retrieval System

1
Freenet A Distributed Anonymous Information
Storage and Retrieval System

• Presenter Chris Grier
• ECE 598nb Spring 2006

2
Outline

• Overview of Freenet
• Architecture and Protocol
• Performance
• Security
• Discussion

3
Freenet Design Goals

• Freenet is basically a P2P file system that
  supports anonymity for readers and writers.
• Anonymity and deniability
• Resistant to denial of service
• Efficient
• Completely Decentralized

4
Keys and Use

• 3 Keys for each file
• Keyword-signed key (descriptive string)
• Signed-subspace key (personal namespace)
• Content-hash key (file contents)
• KSK and SSK (public half only) are published to
  allow retrieval
• CHK allows for updating the file, different CHK
  and same SSK cause collision

5
More Keys and Use

• The structure allows for users to have
  directories.
• Using public/private key pairs allows only users
  who own the space to write files there
• No centralized place to find keys. Must find them
  from someone who publishes.
• Possible crawler
• www publishing of hashes and keys

6
Retrieval

• If the file is not found locally, forwards the
  request on to the node with most similar key.
• Prevents loops by moving down the list if a file
  is not found.
• Hops to live reduces network load and keeps
  requests from progressing too far
• Requests are cached
• Reply message can have the data source changed
  during trip back to node that requested
• Routing Improves over time
• Caching of files replicates popular data closer
  to requests

7
Storing

• Calculate the keys and send an insert message.
  This is very similar to retrieval.
• The hops-to-live here serves to check for
  collisions in the keys generated.
• When a node inserts it checks for a collision
  along the path of insertion.
• The file is stored along the initial path at all
  the hosts
• New files with similar keys are put on hosts with
  other similar keys

8
Management

• Cache uses LRU policy
• Files put in the system might not stay
• All files are stored encrypted, provides
  plausible deniability only
• Discussion is plausible deniability at the node
  operator good enough?

9
Joining the network

• Uses out of band communication for initial join
  msg
• Choose a random seed, send it. Once received the
  next node generates a random seed and XORs with
  the original, and continues the chain.
• After last node generates seed, each host
  releases the seed for verification
• Discussion The structure of the network seems to
  depend somewhat on the initial node chosen. Isnt
  this effected by the out of band communication?

10
Performance and Scalability

• Randomly constructed
• Same cache size and routing table size at every
  node
• The network improves over time as claimed. Path
  length of requests decrease
• As network grows, path length scales like log
• The effect of routing tables can be seen when the
  network gets very large.

11
Security

• Aims to protect readers and writers identity to
  an attack involving collaborating nodes
• No data security, or assurance of data
• Sender Anonymity protected beyond suspicion?
• Without some analysis, this seems like a bold
  statement
• pre-routing suggested, but seems like just
  another idea suggested
• Some aspects of onion routing

12
More Security

• Source protection by randomly flipping the
  data-source field
• Modification of files by attackers is prevented
  by the use of hashes and signatures (CHK and SSK)
• If just stored by KSK, the namespace can be
  dictionary attacked
• Resistant to junk insertion
• Displacement of files requires private keys again.

13
Attacks

• Network construction
• When hosts connect, theres the possibility that
  an attacker can influence how the routes are
  setup.
• Force the network to arrange itself in certain
  ways (this will change)
• Want to determine who sent a message
• With multiple compromised hosts, it seems
  possible might need to guess at topology?
• More attacks?

14
Discussion

• Questions?
• What are some possible attacks on the system?
• Is the lack of search capability acceptable in a
  system like this?
• Does the anonymity provided seem less than hoped
  for in a system like this?
• Does Freenet lack some features you would expect
  in an anonymous publishing/retieval system?
• What does Freenet have to offer over crowds?

15
(No Transcript)
16

• How essential is search?
• A lot of unanswered questions in creating and
  being part of the system
• beyond suspicion for collaberating nodes in the
  sender anonyminty
• Somewhat inspiried by crowds the main difference
  here is the promise of security.
• The authors and readers should be anonymous.
• Scientology documetns being erradicated
• Leaked Diebold source code