In this paper we focus on proposing the effective methods of cyber protection of the application layer. We also discuss how this challenge is related to mission critical industrial and manufacturing systems. In this paper we propose two step HTTP request analysis method that engages request segmentation, statistical analysis of the extracted content and machine learning on the imbalanced data. In this work, we particularly addressed the segmentation technique that allows us to divide the large dataset on smaller subsets and learn the classifiers in a significantly shorter time. In our experiments we evaluated several classifiers that are popular in data mining community. The results of our experiments are obtained on a benchmark CSIC’10 HTTP dataset. The proposed approach allows us to further improve the achieved results of protecting application layer in comparison to other benchmark approaches.