The importance of explainability in machine learning models for cyber security is growing, as it provides a better understanding of the decision-making process and improves the ability to defend against attacks. However, the application of explainability in the context of cyber security is becoming a challenge, as there is currently a lack of standard methodologies for evaluating and comparing the performance and explanations of different models. This paper presents a proposal for a benchmark in the field of explainability of cyberattacks, aimed at enhancing organizations' analysis and response capabilities.