Extension B

Extension B is similar to extension A, but it tolerates packet loss. Extension A does not tolerate packet loss because the receiver needs to know which SEALs the sender discloses so it can update the SEAL boundary. To improve the tolerance to packet loss, we add SEAL boundary information to packets, which increases the communication overhead. This approach allows us to trade off robustness to packet loss with communication overhead.

Two main methods exist to encode the SEAL boundary in packets: absolute encoding or relative encoding. The absolute encoding sends the index of each SEAL of the SEAL boundary in the packet. For instance the SEAL boundary in Figure 4 is $0,\; 2,\; 3,\; 0,\; 1,\; 2$. A relative encoding only communicates the change of the SEAL boundary with respect to a previous boundary.

As in extension A, the receiver needs to know the exact sending rate of messages, to verify that the SEAL boundary always grows by the number of SEALs that the sender discloses. To prevent that an attacker collects more than $r$ SEALs, the receiver needs to receive at least one packet every $\nu$ packets (where $\nu =r/k$). Hence this scheme does not tolerate more than $\nu -1$ consecutively lost packets. Otherwise, an attacker could collect SEALs during a long period of packet loss, and forge subsequent packets by claiming a bogus SEAL boundary.