Light-weight anti-strategic Model for P2P Media-streaming

Ratan Guha, Abeer Hamdy, Darshan Purandare

CoolStreaming, PPLive, etc. are famous chunk based models for live media streaming. However, these systems are prone to different kinds of attacks like pollution attack in which the polluter distributes corrupted chunks to other peers in the swarm. Free riding considers another challenge in P2P file distribution systems in general. Free riders are selfish nodes they join the swarm trying to download without contribution. This selfish behavior leads to degradation in the overall performance of the swarm. This problem emerges mainly from the incentive mechanism used in the P2P system.

Besides, these two kinds of attacks some recent studies on P2P video streaming chunk based models showed some others shortcomings in these models. They showed that:

  • Startup time of video before playback is in order of tens of seconds and sometimes even minutes, and needs to be minimized for a better viewing experience.
  • Some nodes lag in their playback time by minutes as compared to their peers.
  • These streaming services lack tit-for-tat fairness which leads to uneven distribution of uplink bandwidth among users.
  • These approaches use greedy algorithms for peering without the consideration of peer locality that leads to huge cross ISP traffic. Also, random peer selection scheme provides variable streaming quality.

In this project we are planning to propose a P2P media streaming model with security enabled settings inspired by the .torrent file of BT protocols and the power nodes in partially centralized P2P file distribution networks so that the model executes in real time with the desired level of QoS for this type of application. For the proposed model, we need to address the following challenges:

  • Reduce load on the streaming source, start up time before playback, and lag time.
  • Overcome the problem of peers getting media streaming with variable quality due to random peer selection.
  • Use a light weight security techniques to generate and distribute files for detecting attacks in a way that doesn't add much computational and bandwidth overheads to the P2P model.
  • Provide an incentive mechanism in the model to encourage cooperative behavior and punish free riders.