Mining and Generating Vulnerable Patterns for Security Testing

Abstract

Genetic programming is an evolutionary technique that improves individuals from a population in order to better fit the user’s needs. In our case, we apply this technique to create new vulnerabilities from an existing repository mined from the Linux kernel. A first step was to create token sequences from this repository. Then, our genetic algorithm derives new vulnerable patterns according to a fitness function that rely on pattern frequency over the vulnerable dataset. Our results indicate that genetic programming can indeed make vulnerabilities more robust over generations. Our patterns fall into two categories: generic patterns applicable to a large number of files and ones that can
be used on a smaller set.

Date of Award	22 Jun 2018
Original language	English
Awarding Institution	University of Namur
Supervisor	Patrick Heymans (Supervisor) & Gilles Perrouin (Co-Supervisor)