Adversarial attacks have the potential to substantially compromise the security of AI-powered systems and posing high risks especially in the areas like telecommunication where security is a top priority. In this study, we focus on adversarial attacks targeting power allocation for the distributed multiple-input multiple-output networks. We propose a novel defense method to mitigate the effects of these attacks and help boosting the natural performance of the system. The detailed simulations show that the proposed method significantly increases the robustness of the system.