Cadre Formel pour le Test de Robustesse des Protocoles de Communication. (A Framework for Robustness Testing of Communicating Protocols)

In the telecommunication field, protocols have to be seriously validated before their startup. Thus, it is necessary to test the conformance of a protocol, but it is also important to test its robustness in presence of unexpected events. Although a precise definition of robustness is somewhat elusive, functionally the meaning is clear : the ability of a system, conform to its nominal specification, to function in an acceptable way in the presence of hazards. The term "hazards" will be used to gather faults and stressful conditions. The aim of this document is to provide a formal framework for robustness testing for Internet protocols. In order to decide the robustness of an Implementation Under Test (IUT), a clear criterion is needed, taking into account the system behaviors in the presence of hazards. The main contributions of this document are : (1) A framework for robustness testing including an approach, a formal definition of robustness and a test generation method from system specification modelled by an IOLTS. Our approach consists in two methods : The method TRACOR deals with robustness testing in the presence of controllable and representable hazards. It consists in enriching the nominal specification (i.e. protocol standard specification) with invalid inputs, inopportune inputs, acceptable outputs and suspension traces in order to get an increased specification. The method TRACON deals with with robustness testing in the presence of controllable and unrepresentable hazards. It consists in enriching the nominal specification with acceptable outputs and suspension traces in order to obtain a half-increased specification. (2) In order to formalize the robustness of an implementation compared to the increased specification, we propose a binary relation, called Robust, based on the observation of outputs and quiescence of an implementation after executing of traces containing hazards. (3) In order to generate robustness test cases, we propose a specific method based on coloring of the increased specification (or the half-increased specification). This method allows to generate robustness test cases from the increased specification or the half-increased specification using robustness test purposes. (4) The theoretical background of our approach is implemented in the RTCG tool. This application provides three main interfaces, the first one permits to obtain the increased specification, the second one permits to generate robustness test cases using the method TRACOR, and the third one automates the method TRACON. (5) We show how to apply our approach in communicating protocols through a case study on the SSL protocol and TCP.