Machine learning for fuzzing: State of art
Machine learning has become more and more popular in recent years. This popularization has been stimulated by multiple factors: large and affordable computational power, new powerful algorithms, new tools that make it easy to use machine learning algorithms, availability of big data to train the mod...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2020 |
| País: | España |
| Institución: | Universitat Oberta de Catalunya (UOC) |
| Repositorio: | O2, repositorio institucional de la UOC |
| OAI Identifier: | oai:openaccess.uoc.edu:10609/121786 |
| Acceso en línea: | http://hdl.handle.net/10609/121786 |
| Access Level: | acceso abierto |
| Palabra clave: | big data cobertura AFL aprendizaje automático fuzzing coverage machine learning aprenentatge automàtic Machine learning -- TFM Aprenentatge automàtic -- TFM Aprendizaje automático -- TFM |
| Sumario: | Machine learning has become more and more popular in recent years. This popularization has been stimulated by multiple factors: large and affordable computational power, new powerful algorithms, new tools that make it easy to use machine learning algorithms, availability of big data to train the models, etc. Many disciplines have experienced significant changes thanks to its adoption. The fuzzing field has not been an exception. Many researchers have proposed applying machine learning algorithms to the various stages of the fuzzing process. Most studies seem to have brought improvements to the task, however it is not always clear at what cost. Moreover, the reasons behind the selection of one algorithm instead of another are not clear in much of the published literature. This master thesis not only presents the benefits and disadvantages of using various machine learning algorithms in each fuzzing stage, but also identifies new promising paths that researchers should take. |
|---|