Browse Theses

This thesis investigates a recently discovered vulnerability in computer systems which opens the possibility that a single individual with an average user's knowledge could cause widespread damage to information residing in computer networks. This vulnerability is due to a transitive integrity corrupting mechanism called a "computer virus" which causes corrupted information to spread from program to program. Experiments have shown that a virus can spread at an alarmingly rapid rate from user to user, from system to system, and from network to network, even when the best available security techniques are properly used.Formal definitions of self replication, evolution, viruses, and protection mechanisms are used to prove that any system which allows sharing, general functionality, and transitivity of information flow cannot completely prevent viral attack. Computational aspects of viruses are examined, and several undecidable problems are shown. It is demonstrated that a virus may evolve so as to generate any computable sequence, and thus that viruses have Turing capability. Protection mechanisms are explored, and the design of computer networks which prevent both illicit modification and dissemination of information are given. Administration and protection of information networks based on partial orderings are examined, and provably correct automated administrative assistance is introduced. Imprecise viral detection and eradication is examined, and a complexity based integrity maintenance mechanism is suggested. Results of several experiments are given, and the code required to verify results is provided. The relationship of computer viruses to biological, social, and mental systems is explored, and an interpretation of results are given in these domains. (Copies available exclusively from Micrographics Department, Doheny Library, USC, Los Angeles, CA 90089-0182.)