| CPC H04L 9/3278 (2013.01) | 3 Claims |
|
1. A challenge screening method for improving the stability of a strong Physical Unclonable Function, comprises:
Step 1: randomly generating n groups of challenges for testing the stability of a strong Physical Unclonable Function, n being an integer greater than or equal to 100,000; repeatedly inputting each of the n groups of challenges to the strong Physical Unclonable Function m times to obtain m responses corresponding to each group of challenges, m being an integer greater than or equal to 100; if the m responses corresponding to one of the n groups of challenges are identical, determining a first group of challenges as stable challenges, and denoting the stability of the corresponding responses as 1; otherwise, determining the first group of challenges as unstable challenges, and denoting the stability of the corresponding responses as 0, thus obtaining the stability of the responses corresponding to each of the n groups of challenges;
Step 2: defining a second group of challenges and the stability of responses corresponding to the first group of challenges as a challenge-stability pair (CSP); forming one CSP by each of the n groups of challenges and the stability of the corresponding responses, such that n CSPs are obtained, and forming a machine learning training dataset by the n CSPs;
Step 3: obtaining a machining learning model, and adjusting a number of to-be-solved parameters of a machine learning model to be identical with a number of bits of each group of challenges of the strong Physical Unclonable Function, such that each to-be-solved parameter corresponds to one challenge in each group; with the group of challenges in each CSP as an input of the machine learning model and the stability of the responses corresponding to the group of challenges as an output of the machine learning model, training the machine learning model with the n CSPs until the machine learning model is converged, and using the trained machine learning model as a screening model, wherein during a training process of the machine learning model, minimum search is performed on a loss function of the machine learning model through an optimization algorithm, and when the loss function is minimum, it indicates that the machine learning model is converged;
Step 4: during the actual application of the strong Physical Unclonable Function, screening stable challenges by means of the screening model, which specifically comprises:
S1: randomly generating multiple groups of challenges of the strong Physical Unclonable Function, forming an initial challenge set by the multiple groups of challenges, and randomly sorting the multiple groups of challenges in the initial challenge set;
S2: sequentially inputting the multiple groups of challenges in the initial challenge set to the screening model in order, and using an output generated by the screening model when each group of challenges is input as the stability of responses corresponding to this group of challenges, wherein if the output of the screening model is 1, it indicates that the group of challenges input to the screening model are stable challenges; if the output of the screening model is 0, it indicates that the group of challenges input to the screening model are unstable challenges; if a number of obtained stable challenge meets actual application requirements before the multiple groups of challenges in the initial challenge set are not all input to the screening model or after the multiple groups of challenges in the initial challenge set are all input to the screening model, stopping inputting; if the number of obtained stable challenges still does not meet the actual application requirements after the multiple groups of challenges in the initial challenge set are all input to the screening model, returning to S1 until the number of obtained stable challenges meets the actual application requirements; and
Step 5: forming a stable challenge set by all the stable challenges determined in S4, and inputting the stable challenge set to the strong Physical Unclonable Function to extract identity information.
|