The objective of the thesis is to detect the presence of cryptographic primitives in executable binaries. As a part of the thesis, the student should create a dataset of executables for the problem, choose their proper representation, and build the most suitable classifier that will, for the given binary, output a list of cryptographic primitives that the binary contains. Such a classifier will be selected from various models. The dataset creation process should be controlled to allow for adding artificial noise in the form of a non-cryptographic functionality. This is to evaluate the model's ability to generalize and prevent overfitting. The classifier will be trained to recognize at least one cryptographic primitive from the following families: hash functions, symmetric encryption, and asymmetric encryption. The feature vector should be chosen w.r.t. state-of-the-art research on using neural networks for binary analysis. The model shall be subsequently evaluated on an independently obtained dataset. This dataset will come from public repositories or the generator with a different configuration. The thesis should convey whether detecting cryptographic primitives through static analysis and neural network models is feasible.