The model retrieves software behavior information using a hierarchical model of nested auto-associating neural networks

Abstract

The existing algorithms for searching for malicious code are imperfect. For the most part, search programs determine the presence of malicious code based on already known samples. Thus, as a part of the study, a model for extracting information from data was developed, containing hierarchically related basic elements based on deep noise-canceling auto-encoders. In addition, the requirements for the model and the necessary operation parameters were determined.