Algorithm 1 Classifiers testing process.

1:m – user identifier, $m=1,\dots ,M$; 2:$S_i=(sig{n}_i,k)$ – a sample ($sig{n}_i$ – signature, k – claimed class); 3:P – vector of probabilities of belonging $sig{n}_i$ to M classes; 4:$T_a$ – acceptance threshold value; 5:for all samples from the testing set ($i=1,\dots ,K$) do 6:   Select $S_i=(sig{n}_i,k)$. Enter $sig{n}_i$ at the input of the classifier 7:   Check the classifier output P; 8:   if $P\left[k\right]\ge T_a$ and $P\left[k\right]=ma{x}_{m\in M}P$ then 9:       positive user verification; 10:   else 11:     forgery detection; 12:   end if 13: end for