Research And Design Of Intrusion Tolerance Systems By Using Replication

Intrusion tolerance as a new approach to computer security has slowly emerged during the past decade, and gained impressive momentum recently. In this paper, we propose a system based on partial replication what has a lower cost of resource than those based on full replication. In the model, the messages applying for the key services, using the replication protocol, are distributed to the hosts to guarantee the total order, while those applying for the non-key services are distributed to their corresponding hosts directly. However, because local operations on one site are not visible to the others, partial replication may result in inconsistent serialization orders. To solve this problem, a sufficient condition to guarantee the overall correctness is also given-which requires the union of all local serialization graphs to be acyclic.This paper also presents the semi-passive replication technique - a variant of passive replication - that can be implemented in the asynchronous system model without requiring a membership service to agree on a primary.Total order broadcast and multicast is an important problem in distributed systems such as intrusion tolerance. For example, the primary send the update messages and decide messages to the backups by using of total order broadcast and multicast algorithms. So far, there are so many algorithms on total order broadcast and multicast" that have been proposed, following various approaches. It is however difficult to know which solution is best suited a given application context. Though some attempts have been made at classifying and comparing these algorithms, none is comprehensive, and hence lack of generality. So finally in this paper, we propose a classification system based on the mechanisms of ordering. Based on this classification, we define five classes of total order broadcast algorithms, and then relate existing algorithms to those classes.