Modelo de segurança independente de plataforma para execução de software não confiável.

Abstract

Attacks on computer systems have always been a problem and have evolved from simple attacks against physical facilities in the 1970s to coordinated attacks using thousands of computers spread around the world. The main vector of these offensives are softwares that sometimes pass by benign programs but when are installed in a system act in a malicious manner. Environment isolation and behavior detection techniques are used to mitigate the risk of running an unknown and potentially dangerous code. However, many alternatives are expensive and sometimes requires external tools. In this paper, we propose a platformindependent model to provide security to execute untrusted code with no side effects to the host and to third parties. The model consists of two main modules, analyzer and executor, which (a) extracts metadata related to the program and uses them to conduct a preliminary analysis of the code and (b) carry out checks at runtime aimed to preserve the integrity of the system and its associated resources. The validation of the approach was performed by a case study on a volunteer computing application