A study of covert channels in a trusted UNIX system

Abstract

Analysis and identification of potential channels for illicit information flow is not required for Class Bi trusted systems such as the Sun Microsystems Trusted Solaris 1.1 trusted computing base. When used in a multilevel context such channels would present a risk to data security. The problem addressed by this thesis is the identification of covert channels in Trusted Solaris and the determination if their exploitation can be detected using mechanisms provided to the security administrator. The approach taken to address this problem was to identify covert storage channels in the form of observable effects and exceptions of sharing internal databases by subjects at differing access classes. Software was developed to exploit the identified covert channels using a method requiring detailed specifications prior to the creation of code. Audit trails were obtained to evaluate the efficacy of audit in detecting active covert channel exploitation.