FreeNet
A Distributed Anonymous Information
Storage and Retrieval System
Presented By Xiao Wei-Cheng 2004.04.06
Outline
● Introduction ● Architecture
– Keys and searching files
– Inserting, Storing, Retrieving, Managing files – Adding nodes
● Security
Introduction (1/2)
● Decentralization ● Privacy
● Sharing of Storage Space
● Location-independent file system ● Retrieving, Inserting, Storing files
Introduction (2/2)
● Design Goals
– Anonymity for producers and consumers – Deniability for storers of information
– Resistance to attempts to deny access of information – Efficient dynamic storage and routing of information – Decentalization of ALL network functions
Architecture
● Freenet is like a peer-to-peer network ● Files are named by keys
● Each node has its own datastore and routing table ● Routing is driven by 'key'
● Each request has a unique ID – prevent loop
Keys and Searching Files - KSK (1/4)
● Keyword-signed key (KSK) is derived from
Descriptive Text String (DTS)
● public/private key pair
● The file key is yielded by hashing the public part ● The file is encrypted with DTS
● The private part is used to sign the file
Keys and Searching Files–SSK (2/4)
● Signed-subspace key
● Personal namespace is enabled, and generated
randomly
● Files key = hash(XOR(hash(namespace),
hash(DTS)))
● File is encrypted with DTS as KSK
Keys and Searching Files-CHK (3/4)
● Content-hash key
● File key is derived by hashing the file content
● Files are encrypted by randomly-generated keys ● CHK is usually conjucted with SSK
– Indirect file
– Version updating – file splitting
Keys and Searching Files (4/4)
● Problem – How to get the file key ? – Through Web Server
– Lightweight indirect files
● Multiple indirect files may have the same key
Retrieving File (1/2)
● A Request.Data message is sent, with transaction
ID, hops-to-live, depth, and search key
● A Send.Data message and the desired file will be
sent back after successful request
● In the nodes on the path, file is cached, and
routing table is updated
● A Reply.NotFound message would be sent back
if failed
● Files with similar keys would be cached in some
Retrieving File (2/2)
Storing and Inserting File
● A Request.Insert message is sent
● If inserting sucesses, a Reply.Insert message is
sent back, and a Send.Insert is then sent by the requestor
● If failed, a Send.Data message with the existing
data or a Reply.NotFound message is sent back
● In the nodes on the path, file is cached, and
routing table is updated
● New nodes can use inserts to announce their
Managing File
● Storages have finite capacity
● LRU algorithm is used to manage files
● Entries in routing table are deleted only when the
Adding Nodes
● All nodes have to be consistent in deciding the
new node key (Address Resolution Key)
● The new node sends its address and hash(rand())
out first
● Nodes in the path send hash(rand() XOR prehash)
to the next one
● The final hash value becomes the key of the new
Security (1/3)
● Anonymity of sender, receiver, and the key
● Key anonymity is impossible since routing
depends on the key
● For malicious nodes, sender anonymity is
Security (2/3)
● Freenet + pre-routing
– For key anonymity and sender anonymity
– Messages are encrypted by a succession of public
keys, and pre-routed first
– After pre-routing, the message is injected into the
normal Freenet network
● The data source field can be resetted in the path ● A hops-to-live of 1 doesn't reveal an endpoint
Security (3/3)
● Modification of requested files by malicious nodes – Not feasible under CHK or SSK
● Displace existing files by malicious nodes – Not feasible under CHK or SSK
● Prevent DoS attack
– Use 2 part of datastore
● Established files ● New files
Performance Analysis (1/5)
● Network for simulation – 1000 nodes
– Datastore size of 50 items per node
Performance Analysis (2/5)
● Network Convergence
Performance Analysis (3/5)
● Scalability
