Towards Automated Proving of Relational Properties of Probabilistic Programs (Invited Talk)
Some security properties go beyond what is expressible in terms of anindividual execution of a single program. In particular, many securitypolicies in cryptography can be naturally phrased as relationalproperties of two open probabilistic programs. Writing and verifyingproofs of such properties is an error-prone task that calls forautomation and tool support. One of the main difficulties inautomating these proofs lies in finding adequate relational invariantsfor loops and specifications for program holes.In this talk we showhow to automate proofs of relational properties of open probabilisticprograms by adapting techniques for the automatic generation ofuniversally quantified invariants in a non-relational setting.