Reducing Damage Assessment Latency in Survivable Databases

Abstract

Traditional recovery mechanisms are not adequate in protecting databases from malicious attacks. A malicious transaction by virtue of writing on to the database can corrupt one or more data items; benign transactions reading these data items and writing on other data items can help spread the damage. To prevent the damage from spreading, it is important to assess the damage and confine it as quickly as possible. Algorithms providing fast damage assessment are needed. In this paper we look at two existing techniques for damage assessment and analyze their complexity. We also propose a new technique that improves upon the existing techniques by reducing the time required for damage assessment.