Abstract Certification of Java Programs in Rewriting Logic

  1. Alba Castro, Mauricio Fernando
Supervised by:
  1. María Alpuente Frasnedo Director

Defence university: Universitat Politècnica de València

Fecha de defensa: 23 November 2011

Committee:
  1. Salvador Lucas Alba Chair
  2. Alicia Villanueva García Secretary
  3. Ricardo Peña Marí Committee member
  4. Francisco Durán Muñoz Committee member
  5. Claude Marché Committee member

Type: Thesis

Abstract

In this thesis we propose an abstraction based certification technique for Java programs which is based on rewriting logic, a very general logical and semantic framework efficiently implemented in the functional programming language Maude. We focus on safety properties, i.e. properties of a system that are defined in terms of certain events not happening, which we characterize as unreachability problems in rewriting logic. The safety policy is expressed in the style of JML, a standard property specification language for Java modules. In order to provide a decision procedure, we enforce finite-state models of programs by using abstract interpretation. Starting from a specification of the Java semantics written in Maude, we develop an abstraction based, finite-state operational semantics also written in Maude which is appropriate for program verification. As a by-product of the verification based on abstraction, a dependable safety certificate is delivered which consists of a set of rewriting proofs that can be easily checked by the code consumer by using a standard rewriting logic engine. The abstraction based proof-carrying code technique, called JavaPCC, has been implemented and successfully tested on several examples, which demonstrate the feasibility of our approach. We analyse local properties of Java methods: i.e. properties of methods regarding their parameters and results. We also study global confidentiality properties of complete Java classes, by initially considering non--interference and, then, erasure with and without non--interference. Non--interference is a semantic program property that assigns confidentiality levels to data objects and prevents illicit information flows from occurring from high to low security levels. In this thesis, we present a novel security model for global non--interference which approximates non--interference as a safety property.