The development of technologies and the Internet conditioned the development of video surveillance systems. In modern video surveillance systems, the complete content recorded by a camera can be publicly available, i.e., visible on the Internet. The mass use of this type of camera has led to the fact that the cameras record everything and everyone and that the recorded content violates privacy. In order to protect privacy, different types of technologies and software are applied. One of those technologies is machine learning. By applying machine learning, it is possible to mask people's faces on recorded material. In this way, in publicly available content, as well as in case of unauthorized access, the identity of people is not compromised. This paper describes the architecture of a system that enables the application of machine learning for the purpose of protecting the identity of a person, where only an authorized person would have access to the original unmasked content.