salle A. Turing CE4
2 juin 2014 - 15h30
On some ideas for vulnerability discovery using Machine Learning
par Gustavo Grieco de Université de Rosario (Argentina) et Verimag
Abstract: A typical OS usually involves thousands of executable files.
For these programs, we can have thousands of bugs reports.
Fuzzing and detecting vulnerabilities at the OS scale leads to
new issues and challenges. We think can address these using
a combination of dynamic/static program analysis with Machine Learning
techniques in order to learn pattern to discover vulnerable programs
and guide the fuzzing process to generate
In this short talk, we will present some ideas and preliminary results
on how to collect, process and learn from data in order to discover
security vulnerabilities directly on binary programs.
