Year of Award

2012

Document Type

Thesis

Degree Type

Master of Science (MS)

Degree Name

Computer Science

Department or School/College

Department of Computer Science

Committee Chair

Mike Rosulek

Keywords

cryptography, functional encryption, obfuscation, theoretical computer science

Publisher

University of Montana

Abstract

We introduce a new model for program obfuscation, called mediated obfuscation. A mediated obfuscation is a 3-party protocol for evaluating an obfuscated program that requires minimal interaction and limited trust. The party who originally supplies the obfuscated program need not be online when the client wants to evaluate the program. A semi-trusted third-party mediator allows the client to evaluate the program, while learning nothing about the obfuscated program or the client’s inputs and outputs. Mediated obfuscation would provide the ability for a software vendor to safely outsource the less savory aspects (like accounting of usage statistics, and remaining online to facilitate access) of “renting out” access to proprietary software. We give security definitions for this new obfuscation paradigm, and then present a simple and generic construction based on functional encryption. If a functional encryption scheme supports decryption functionality F (m, k), then our construction yields a mediated obfuscation of the class of functions {F (m, ·) | m}. In our construction, the interaction between the client and the mediator is minimal (much more efficient than a general- purpose multi-party computation protocol). Instantiating with existing FE constructions, we achieve obfuscation for point-functions with output (under a strong “virtual black-box” notion of security), and a general feasibility result for obfuscating conjunctive normal form and disjunctive normal form formulae (under a weaker “semantic” notion of security). Finally, we use mediated obfuscation to illustrate a connection between worst-case and average-case static obfuscation. In short, an average-case (static) obfuscation of some component of a suitable functional encryption scheme yields a worst-case (static) obfuscation for a related class of functions. We use this connection to demonstrate new impossibility results for average-case (static) obfuscation.

Share

COinS
 

© Copyright 2012 Robert Perry Hooker