Numeric Program Analysis Techniques with Applications to Array Analysis and Library Summarization
Abstract
Numeric program analysis is of great importance for the areas of software engineering, software verification, and security: to identify many program errors, such as out-of-bounds array accesses and integer overflows, which constitute the lions share of security vulnerabilities reported by CERT, an analyzer needs to establish numeric properties of program variables. Many important program analyses, such as low-level code analysis, memory-cleanness analysis, and shape analysis, rely in some ways on numeric-program-analysis techniques. However, existing numeric abstractions are complex (numeric abstract domains are typically non-distributive, and form infinite-height lattices); thus, obtaining precise numeric-analysis results is by no means a trivial undertaking. In this thesis, we develop a suite of techniques with the common goal of improving the precision and applicability of numeric program analysis. The techniques address various aspects of numeric analysis, such as handling dynamically-allocated memory, dealing with programs that manipulate arrays, improving the precision of extrapolation (widening), and performing interprocedural analysis. The techniques use existing numeric abstractions as building blocks. The communication with existing abstractions is done strictly through a generic abstract-domain interface. The abstractions constructed by our techniques also expose that same interface, and thus, are compatible with existing analysis engines. As a result, our techniques are independent from specific abstractions and specific analysis engines, can be easily incorporated into existing program-analysis tools, and should be readily compatible with new abstractions to be introduced in the future.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2007
- Accession Number
- AD1165880
Entities
People
- Denis Gopan
Organizations
- University of Wisconsin–Madison