Fig. 4: Computer-vision-based authentication/identification of artificial fingerprint PUFs.

a Example of a successful authentication/identification and (b) example of a non-successful authentication/identification. These examples show the process of overlapping the test image on the database one through homographic transformation, where the overlapped image obtained through XOR operation is generated to graphically evaluate the overlapping of images under consideration. When images of the same nanopattern are compared (a), the XOR overlapped image is composed mainly of “0” values since the homographic transformation enables to correctly overlap the test image on the top of the database one. When images of two nanopatterns are compared (b), the XOR overlapped image is constituted of almost equal “0” and “1” values”. Indeed, in this latter case, the homographic transformation results in a strong distortion of the test image while trying to overlap with the database one. In (a, b) the colored lines connecting the test image with the database image represent the matches of the homographic transformation, in fact, they connect corresponding key points (represented by circles) in the two images. Dashed lines in the XOR overlapped images represent the area of effective image overlapping, i.e., the area where the test image is effectively overlapped with the database one. Heat-map matrix representing the fractional HD (c) and corresponding intra and inter-distributions between images from the test set and the database (d) images acquired after 6 months and the images from database to test aging stability (e, f) images acquired after thermal annealing of PUF devices at 200 °C for 30 min and images from the database to test high-temperature thermal stability (g, h) and images acquired after thermal annealing of PUF devices at −196 °C for 30 min and images from the database to test low-temperature thermal stability (i, j). Dashed lines in intra and inter-distributions represent the decision threshold evaluated from the comparison between the test set and the database.