Abstract

The secure and efficient detection of process failures is an essential requirement of many distributed systems. In this paper, we present the design and analysis of a mechanism used for the detection of member failures in secure groups. Based on one-time passwords, our solution does not obviate the need for periodic statements from group members, but significantly reduces the cost of their generation and validation. A study comparing the costs of traditional mechanisms with our proposed approach is presented. Results of the study indicate the average case performance of the proposed scheme is ¤¦¥§¤©¨���� of traditional failure detection in trusted groups, and negligible in the untrusted groups. A discussion of security and performance tradeoffs made through mechanism policy is provided. 1

Citations