Secure atomic multicast primitives for distributed information assurance applications

Distributed agreement protocols among a group of application processes are often built upon atomic multicast message delivery guarantees. Issues arise however when atomic multicast is employed in information assurance (IA) settings where extreme failure behaviors such as send-omission of processes and message timeliness violations are likely. The issues are compounded by security weaknesses in the communication software that make it easier for intruders to stage attacks. These issues impact the design of atomic multicast protocols, which have hitherto assumed only benign failures such as process crash and network message loss/delay. In this paper, we revisit atomic multicast protocols, taking into account the IA dimension as well. Our study reveals the need for a secure centralized entity to realize the multicast message delivery functions. Two functions are studied: dynamic membership management to exclude nonresponsive members from the group and security-reinforced communications to enforce mutual anonymity of members. Our paper walks through the design of a secure atomic multicast protocol and its use in IA applications.