Deep Kamal Singh
HostClient sends a request to Tracker.
Tracker- Provides .torrent file containing
information of the other clients.
Clients - Performs both Upload and Download.
HostClient communicates with the clients
received in .torrent.
Clients exchange pieces (different strategies) &
match their hashes with those received in .torrent.
Samsara: “Honor among thieves”
Cheating strategies in BitTorrent
◦ Under-reporting strategy
◦ Collusion in BitTorrent
Using the trust relationships to eliminate a trusted
Restrict unwanted communication from “Sybil-
User with multiple identities can send additional
unwanted communication by making additional
Every client in the swarm gets a “Fair-share”
1. A trusted entity (Tracker) observes all actions for
every client and associates them with the identity of
the Client performing the action
2. Every Client classifies the received packets of data
as wanted or unwanted.
3. There is a non-trivial cost for initiating and
maintaining links in the network.
All the transactions consists of three phases.
Authorization -Tracker checks if the transaction can be
approved and a token could be issued to the sender.
Transmission – As per the token, sender takes the
Classification – Receiving client classifies the
communication (Good or Bad) & provides feedback to
Use link credit instead of user credit.
Directional credit design
Each link with a link credit balance B, initial
value 0, and range [L, U]. With L <= 0 <= U,
and L<= B <= U.
A client before a Initial credit of 20 Mb is
transaction requests the assigned to every link
Tracker for approval. Receives a request for
On the reply, the transaction from sender.
transaction is declined or Issues a token for the
After the receiver’s
classification, makes the
At Client At Tracker
Simulated on 80 Planet-Lab hosts all around the
world (MA, NY, CA, India, Canada, Hong Kong
China, Taiwan etc.).
Tracker – CCIS Lab Machine (God save TimePilot !!).
a. 924 MB
b. 693 MB
c. 81 MB
Performance of original snark
Credit-based Tracker, without sybil nodes in
Comparison with sybil nodes in snark
Credit-based Tracker, with sybil nodes in
Implementing the system using a decentralized
Using the social network to gather the torrent
Developing and implementing a new Incentive
mechanism for untrustworthy clients.
Data is collected and plotted for all nodes