Knowledge-guided countermeasure attack and defense method of deep learning soft measurement model

The invention discloses a knowledge-guided adversarial attack and defense method for a deep learning soft measurement model, and the method comprises the following steps: firstly, building an adversarial attack frame for the deep learning soft measurement model, and proposing the characteristics which are required to be satisfied for the attack to have the feasibility; then, proposing a knowledge-guided adversarial attack method to attack the deep learning soft measurement model, introducing mechanism knowledge into a target function, adding a new constraint condition, and defining a barrier function to reconstruct an optimization problem so as to solve the ill-conditioned condition of the optimization problem; and finally, on the basis of the provided attack method, providing a corresponding knowledge-guided adversarial training defense method, and carrying out active defense in combination with an adversarial sample and an initial deep learning soft measurement model. Compared with other adversarial attack