Detection and analysis of network security threats using artificial intelligence

This article describes ways of detecting and analysing major security threats in network. Our research is based on a neural network created in Python language with the use of TensorFlow. With the help of artificial intelligence, we can analyse language for the presence of disinformation and propaganda. We have used BERT (Bidirectional Encoder Representations from Transformers) model, in order to catch fake, malicious and suspicious content. Basing on the BERT model, other models were created, for example HerBERT, which was trained to understand Polish. Furthermore, we have availed ourselves of DNS analysis, so we can localize IP addresses of Internet domains. In this work, we will discuss these methods of protecting information space.