Abstract: A computer-implemented method or protecting a machine-learning model against training data attacks is disclosed. The method comprises performing an initial training of a machine-learning system with controlled training data, thereby building a trained initial machine-learning model and identifying high-impact training data from a larger training data set than in the controlled training data, wherein the identified individual training data have an impact on a training cycle of the training of machine-learning model, wherein the impact is larger than a predefined impact threshold value. The method also comprises building an artificial pseudo-malicious training data set from the identified high-impact training data and retraining the machine-learning system comprising the trained initial machine-learning model using the artificial pseudo-malicious training data set.

Abstract: A riblet film has a riblet structure on a first side of the riblet film; and a fixing surface on a second side of the riblet film. The riblet film is formed from a cured embossing lacquer, in which a planar textile is embedded, and in which the riblet structure is embossed.

Abstract: A truncated cone is disclosed for molding bottoms of plastic containers, such as PET bottles, into geometries which extend convexly inwardly, are thermostable during hot-washing, and allow substantially full visual inspection with minimum shadow effects. The truncated cone is in the form of a frustoconical bottom insert, which, in a first blow-molding step, creates an intermediate container with a bottom geometry having a substantially wavy cross section. The wavy cross section of the bottom geometry can be accentuated by means of a second mold with a plateau, which creates a slight, upper domed surface in the bottom geometry of the container molded thereby.