Handling Undiscovered Vulnerabilities Using a Provenance Network

Abstract

This paper elaborates on a novel approach at preventing exploits from vulnerabilities which remain uncovered during the testing phase of a system's development lifecycle. The combination of predicted usage patterns, a Provenance network model and a clustering methodology provide a secure failure mechanism for both known and unknown security issues within the system. The paper also addresses of the requisite supporting infrastructure and deployment issues related to the model. The idea is to approach the growing problem of newer and more complex vulnerabilities in an ever more intricate and vast set of systems using a generic software state mapping procedure for recognizable (and thus the complementary unrecognizable) patterns to judge the stability at each step in an operation sequence. Thus abstracting these vulnerabilities at a higher level provides us a generic technique to classify and handle such concerns in the future and in turn prevent exploits before a corrective patch is released.