next up previous
Next: Dealing with clock drift. Up: Initial Synchronization - Further Previous: Initial Synchronization - Further

Combining with multicast group control centers.

The general IP multicast model assumes that any host can join the multicast group, receive all group data, and send data to the group [11]. To join the multicast group, the receiver only needs to announce its interest to a local router which takes care of forwarding packets to that receiver. Each joining group member contacts a central server or a group controller to negotiate access rights and session keys. This model is supported by the Secure Multicast Users Group (SMUG) [29] and we adopt it for our secure authentication scheme, which requires that each receiver performs an initial registration (for time synchronization and interval timing information) at the sender or at a central server.

Here is a sketch of a scalable synchronization mechanism that uses this infrastructure: Both senders and receivers synchronize with time synchronization servers which are dispersed in the network. After the synchronization, every entity E knows the time and the maximum error δt(E). The sender S periodically broadcasts a signed message which contains δt(S), along with the interval and key chain commitment information for each authentication chain. A new receiver R therefore only need wait for the broadcast packet allowing it to compute the synchronization error between itself and the sender as δt = δt(S) + δt(R). Based on the δt the receiver determines the minimum-delay authentication chain it can use. Hence, the receiver does not need to send any messages to the sender, provided that the sender and receiver have a method to synchronize and the receiver knows the upper bound of the synchronization error δt.


next up previous
Next: Dealing with clock drift. Up: Initial Synchronization - Further Previous: Initial Synchronization - Further

Adrian Perrig
Sat Sep 2 17:01:14 PDT 2000