P2P Taxonomy and Security Concerns Ryan Lackey CTO, HavenCo, Ltd. RSA Conference 2002 San Jose
Introduction Variety of P2P systems P2P is not a new thing
Types of Systems “Traditional” p2p File Sharing Proxies Chat systems Infrastructure systems
Major File Systems Napster Gnutella KaZaA/Fasttrack/Morpheus Freenet Mojonation
Traditional SMTP is peer to peer Deployed with “supernodes” with smtp/pop3 and inter-realm communication via supernodes
Cypherpunks-style r ers 35 or so nodes “Onion routing”
Chat Systems IRC isn’t really p2p AIM/ICQ with centralized presence Gale, Jabber, IMPP proposals
Infrastructure Protocols DNS NTP PKI Certification Authorities
Design Comparison Target applications Transport Interactivity Degree of centralization Design/compile-time organization or install/configuration or runtime/evolving Security: traffic encryption, DoS protection, Replication for reliability
Implementation Comparison “Official” vs. covert adoption Importance of “network effects” for minimal utility Legal issues (content, copyright controls) Administrative control – what functionality is possible, who exercises it?
Security Issues Users provided an incentive to violate security model System not designed to be compatible with non-P2P restrictions Modifies underlying assumptions about connectivity
Observations “Old” p2p systems ( , etc.) seem to be designed into security models, so newer systems can emulate Power ultimately wins over security Systems can be re-deployed internally for security
Summary Since P2P applications have been popular, and continue to be popular, security practices must take them into account Deployment choices are as important as implementation choices; even unsafe technologies can be wrapped in a security model
Q&A