Satellite operators rely on space situational awareness (SSA) data to avoid collisions with other space objects. However, SSA data can also be used by adversaries to launch both kinetic and non-kinetic attacks against satellites. An existing approach to prevent these attacks consists in labeling a sensitive satellite as a debris in the SSA data. Its objective is to hide the satellite's true label from potential adversaries. Unfortunately, adversaries can easily determine the true label of a satellite based on the orbit information provided by the SSA system. In this paper, we propose a novel adversarial machine learning approach to conceal sensitive satellites in SSA data from potential adversaries. Our method conceals the satellite by changing both its label and the orbit information. The resulting SSA data entry cannot be used to determine that the space object has been mislabeled. Our experimental results show that our proposed method reduces the ability of an adversary to determine the true label a of a concealed satellite to less than 50%. © 2024 IEEE.