A critique of the DeepSec Platform for Security Analysis of Deep Learning Models