Having Fun with P2P Keith W. Ross Polytechnic University ross.

Having Fun with P2P

Keith W. RossPolytechnic University

http://cis.poly.edu/~ross

Many Apps Migrating Client-Server to P2P

• File sharing• File and patch distribution• Live video streaming• Video on demand• VoIP• Hybrid CDN/P2P

Today’s Talk

• P2P security– Is BitTorrent Unstoppable?

• Survey of P2P video streaming research– Creating an open P2P video streaming

ecosystemResearch supported by: NSF, Microsoft, Movie Labs, Huawei, Verizon, Panasonic, Len Shustek

Why study P2P Security?

• P2P is potentially more vulnerable than client server.

• Need to understand the security issues for architecting future P2P apps

• Attacks from entertainment industry reveal weak spots in P2P

Security work joint with:

Jian Liang Rakesh Kumar

Prithula Dhungel Di Wu

Naoum Naoumov

Xiaojun Hei

Earlier work onKazaa and eDonkey

• FastTrack/Kazaa– Unstructured P2P network

• Overnet/Kad– Structured (DHT) P2P network– Part of eDonkey/eMule

File Pollution: Infocom 05

pollution company

polluted content

original content

File Pollution

pollution company

pollution server

pollution server

pollution server

pollution server

file sharingnetwork

File Pollution

Unsuspecting usersspread pollution !

File Pollution

Unsuspecting usersspread pollution !

Yuck

Index Poisoning: Infocom 06

index title location bigparty 123.12.7.98smallfun 23.123.78.6heyhey 234.8.89.20

file sharingnetwork

123.12.7.98

23.123.78.6

234.8.89.20

Index Poisoning

index title location bigparty 123.12.7.98smallfun 23.123.78.6heyhey 234.8.89.20123.12.7.98

23.123.78.6

234.8.89.20

index title location bigparty 123.12.7.98smallfun 23.123.78.6heyhey 234.8.89.20bighit 111.22.22.22

111.22.22.22

FastTrack Copies

Overnet Copies

Is BitTorrent Unstoppable?

(IPTPS 2008; extended version available)

• How can record/movie companies limit piracy in BitTorrent?– Suing companies?– Suing users?– Internet attacks?

• How vulnerable is BitTorrent to attacks?

BitTorrent

tracker: tracks peers in torrent; providestracker list

torrent: group of peers exchanging chunks of a file

trading chunks

peertorrent index server: search for torrents;provides .torrent file

BitTorrent Ecosystem

• Open protocol– 50+ client implementations– Dozens of tracker implementations– Dozens of torrent location sites

• 5 million simultaneous users & growing

• Evolving:– Peer discovery: DHTs, gossiping– Proprietary protocols, private torrents

BitTorrent Basics

• Seeds and leechers• File divided into 256KB pieces. Each piece

is 16 blocks.– Download blocks and assemble pieces– Hash piece to check integrity

• Peers advertise pieces they have to neighbors

• Peer sends blocks to four neighbors currently sending it data at the highest rate

• And also to one random neighbor

Classes of BitTorrent Attacks

Attacks against an existing torrent– against leechers– against initial seed– against peer discovery– against peer discover

Decoy attacks: attacker creates own torrent– Seeding a polluted file– Seeding a file and delivering only 99%

Fake Block Attack

• Attacker establishes TCP connections with legitimate peers

• Peer downloads one fake block from attacker – and 15 good blocks from legit peers– Hash failure – download is prolonged

Acknowledgment: Thanks to Vishal Misra for bringing this attack to our attention.

Simple analysis of fake block attack

What is the probability of a hash failure?

• n = # neighbors advertising piece• m = attack nodes (m < n)• k = # neighbors from which it is

downloading the 16 blocks

Probability of a clean piece

• 70% failure requires 20% attack neighbors

• γ increases for rare pieces and in end game

nm /

Connection attack

• Attacker establishes many TCP connections to each target peer.– Doesn’t upload any blocks– Chatty peer: keeps connection active

with repeated BT handshake messages

Passive Measurements

• Collect traces while downloading– Azureus and uTorrent– DSL and Ethernet– 54 downloads of “Foo Fighters” (108 MB)

• Developed parser to analyze BT trace• To estimate download time without attack:

– Obtain blacklist from torrentfreak.com– Use Peer Guardian to prevent connections to

blacklisted peers

Azureus results

• Download is NOT being prolonged by more than 50% for DSL

Zoom in one Azureus trace

Chatty-peers make upa major fraction of the useful peers.

Handshake messages sent by chatty peers

uTorrent

Zoom in on uTorrent trace

Passive Measurement Conclusions

• Anti-P2P companies applying different strategies for different BT clients

• Largely ineffective for Ethernet clients

• For DSL, download time increases by 30-60%

Active Measurements

• Crawl peers in torrent and identify– Fake-block attack peers– Chatty peers

• Looked at 8 box-office movie torrents– Some under attack; some not

• Crawler speaks Azureus protocol to Azureus peers – and conventional protocol to all other

peers

Active Measurement Conclusions

• Several, but not all, top-box movies are under attack

• Published blacklists do not cover all the attackers in a torrent

• Most attackers enter through gossiping

Seed Attack

• “Nip in the bud”

• Make many connections to seed, download at high rate

• Rationale: Conventional algo gives all its bandwidth to 5 highest downloaders

Planet Lab experiments

• Put 30 leechers on PL nodes; upload capacity capped at 512 kbps

• Seed upload capacity 160 kbps – Azureus and uTorrent

• 0-60 attack peers

• Start seed, start 5 leechers, start attack peers, start 25 leechers

Seed attack results

delay attacker bandwidth

BitTorrent Conclusions

• Attacks against leechers can prolong downloads– But is 50% enough?– Blacklists are helpful; adaptive blacklist is

needed• Seed is surprisingly resilient to attacks• Attacks must be tailored to client types

• BitTorrent ecosystem is difficult to stop!

Today’s Talk

• P2P security– Is BitTorrent Unstoppable?

• Survey of P2P video streaming research– Creating an open P2P video streaming

ecosystemResearch supported by: NSF, Microsoft, Movie Labs, Huawei, Verizon, Panasonic, Len Shustek

Live Streaming: Faculty Collaborators

ProfessorYong Liu

ProfessorYao Wang

ProfessorShiv Panwar

and Professor Nitsesh Saxena (CS)

obtain listof peers

trade chunks

tracker

peer

Basic idea P2P live streaming

Sourceof video

PPLive: Chinese New Year 2006

@ Polytechnic

• Infrastructure for measuring P2P live video systems– Xiaojun Hei, Jian Liang, Yong Liu

• Stochastic modeling of P2P live streaming– Rakesh Kumar, Yong Liu

• Incentives and scalable video for live P2P streaming– Zhengye Liu, Yanming Shen, Shiv Panwar, Yao Wang

• Security of P2P live streaming systems– Prithula Dhungel, Xiaojun Hei, Nitesh Saxena

• P2P VoD, Hybrid P2P-CDN– Angela Wang, MSR colleagues

• Developing our own live P2P video system– Master’s students

Open P2P Live Streaming

• Create an ecosystem for live streaming• Need incentives

– The more you upload the better the quality

• Layered video– Chunk-based– Substream-based

• Accommodate user generated content

Thank You !