Privacy is a fundamental human right and should be absolutely respected. It implies that in the area of monitoring systems, anonymization methods should be safe and robust. This study summarizes our experiments and results related to the IEEE BigData 2022 Cup: Privacy-preserving Matching of Encrypted Images data mining competition. Its goal was to verify three different obfuscation schemes. In the case of two subtasks, our predictive model was built on image statistics, such as average color, its standard deviation, or interquartile range. The algorithm achieved 99.88% and 95.45% accuracy, respectively. In addition, we proposed a novel, to our knowledge, data augmentation approach that improved the results very slightly but, in our opinion, has potential and is worth further analysis. The third obfuscation scheme, based on homomorphic encryption, turned out to be beyond our reach. In the end of paper, ideas for further research are provided.