Skip to main content

Table 2 Accuracy on the Dresden dataset in the closed-set scenario without compression (random-choice accuracy is 0.016)

From: Combining PRNU and noiseprint for robust and efficient device source identification

  1. Here and in next tables, the best fusion results are in boldface. When they improve by 20% or more over PRNU, they are emphasized with a green background. Large improvements are observed when there is data scarcity (small d, small N). All classifiers have quite similar performances