Tor guard musings

This is not a serious proposal– I just couldn’t sleep and was thinking about the protocol design.

Using a redesigned BitTorrent protocol, Tor clients will distribute traffic amongst at-least-two guard relays. Guard relays will only handle session fragments. Second hop relays will be responsible for recombining and splitting all traffic sent to and from the client.

Because guard relays will only process partial Tor traffic and the client and second hop relay is responsible for recombining and splitting traffic, it is feasible for second hop relays to transmit session fragments to randomized set of guard relays.

Second hop relays could intelligently pass BitTorrent traffic to any number of guard relays, dynamically, instead of relying on a static circuit.

Advantages

  • Guard nodes never receive 100% traffic.
  • All relays can be guard relays 24/7/365.
  • Network latency increases because the client and second hop relay will have to wait for all packets for any given session.

Disadvantages

  • Relays will require added CPU processing.
  • Network is responsible for processing greater amounts of data.
  • Network latency increases because the client and second hop relay will have to wait for all packets for any given session.