Printing and scanning investigation for image counter forensics

EURASIP Journal on Image and Video Processing

Table 3 Validation accuracy for various validation sets after training on IFSTC data set

We note that although all three models perform exceptionally well on the original IFSTC data set, each performs little better than random when evaluated on images from any of the three printers. Because JPEG compression and Bilinear Resampling (RS) could be reasonably inferred to be similar to printing and scanning, we remove these classes and train and evaluate on a restricted set of four classes (4c) (see Sect. 4). Despite this restricted set of manipulations, however, the models perform no better than random