<ul><li>A novel framework is proposed for ownership verification of deep neural network (DNN) models for image classification tasks.</li><li>It allows verification of model identity without presenting the original model, suitable for scenarios where an unauthorized user has a copied model in a cloud environment.</li><li>The framework uses a white-box adversarial attack to align output probabilities, enabling the rightful owner to identify the model.</li><li>The proposed method based on the iterative Fast Gradient Sign Method (FGSM) with control parameters shows effective identification of DNN models using adversarial attacks.</li></ul>

Ownership Verification of DNN Models Using White-Box Adversarial Attacks with Specified Probability Manipulation

Discover more