Palestra: Android Permission's Analysis Using Ontologies and Machine Learning

Data: 
21/10/2015 - 14:00
Local: 
Sala 85 - IC-2
==================================================================
            TODOS SÃO BEM-VINDOS! 
==================================================================

                Universidade Estadual de Campinas - UNICAMP
                             Instituto de Computação - IC
                Laboratório de Segurança e Criptografia (LASCA)

 
                            PALESTRA LASCA 2015
                   Quarta-feira, 21/10/2015, às 14:00 horas
                                       Sala 85 - IC-2
 
               ANDROID PERMISSION'S ANALYSIS USING
               ONTOLOGIES AND MACHINE LEARNING
 
 
                              Luiz Claudio Navarro
                       (lcnavarro@lasca.ic.unicamp.br)
 
 
                                    ABSTRACT
 
Success on analysis and construction of any system depends on a good model which allows to represent components, properties, relationships and consequently features and behaviors. As ontologies are the way to represent data with their associated meaning (semantics) and inference logic, it is a good foundation for creating system models which are computable. They can be stored and processed using ontology technologies as defined on Web Semantic arena. In the other hand, a system description easily grows in complexity up to levels where it is difficult to establish cause-effect inference rules due to the complicated chain of conditions that are normally found in actual models. Then machine learning can help to identify what are the elements, properties and relations which are important on the cause-effect complex relationship. It is not the final solution as it depends on a large number of samples and the quality of information captured, but it can help on the process of investigation and model refinement. Particularly on the malware attacks, permissions have an important role as they are part of the resources' protection mechanism . This presentation shows a method to represent Android packages and permissions using OWL (Ontology Web Language) and a way to analyze the model using Random Forest as a base algorithm of a machine learning approach to identify permissions which can be related to malwares.
 
 
===============================
Responsável: Marcelo I.P. Salas
LASCA, Sala 84
Instituto de Computação, UNICAMP
===============================