Dynamic Detection of Program Invariants using Genetic Programming

Dynamic Detection of Program Invariants using Genetic Programming
Luigi Cardamone
PhD Student

DEI - Sala Seminari
4 novembre 2010
Ore 11.30

Abstract:

Software testing is a critical phase of the development process and can be very complex when the software system is wide.
Hence, there is the need of providing the programmer with automatic tools able to infer some high level properties of the program to be tested.
The extracted information can be used either to speed up the debugging phase or to prove the reliability of the software.
In recent years, different approaches have been proposed to extract interesting information about a given program.
Among these, one promising approach is the tool DySy which is based on a dynamic symbolic execution technique.
DySy exercises the functionality of an application with the help of a test suite and simultaneously also performs a symbolic execution of the program to infer properties without any predefined set of patterns.
In this work we present our approach that aims to improve the inference of properties in body loops. Our approach use an evolutionary algorithm-based methodology, Genetic Programming, to search for a logical formula that summarizes the computation of a loop. The results show that our approach is very effective and can be integrated in a tool like DySy to improve the overall performance.

Contatti:
Luigi Cardamone

Area di ricerca:
Intelligenza artificiale, robotica e computer vision


PhD Student

DEI - Sala Seminari
4 novembre 2010
Ore 11.30