... information implicit in the models against model inversion attacks. Hence, this paper presents the development of a new approach called mosaic neuron perturbation (MNP) to preserve latent information in the framework of the AI model, ensuring differential privacy requirements while mitigating the risk...