An EFSM-based model for testing security issues in the Hadoop ecosystem

Abstract

Today, the Hadoop infrastructure can be affected by several forms of attacks on public or hybrid clouds. In this work, we propose a comprehensive security scheme based on Kerberos and Adaptive eXtensible Access Control Markup Language (XACML) by considering some timing constraints. In a related context, the appropriate approach to checking the vulnerability of a Hadoop implementation is to perform tests according to predefined specifications. Therefore, the second contribution is to model the security system specifications using an extended finite state machine (EFSM) and to propose an architecture to handle the Hadoop testing process. Furthermore, we provide an algorithm to generate secure local test sequences (SeLTS) that describe the test execution at each port of the system. Finally, experimental results come along to validate our scheme in the healthcare system. As a result, we notice a remarkable time decrease in access request processing with a performance improvement using the new time-constrained approach.