Named Entity Recognition Methods in Zero Knowledge Network Traffic Analysis

The article discusses methods for searching address fields in network traffic without a priori information about its structure using named entity search methods from natural language processing algorithms. The urgency of the problem being solved is explained, due to the complexity in meeting the requirements of the law in ensuring information security in the automated control system and IoT networks. The method of generating the initial data, the method of its preprocessing (calculation of statistical characteristics) to improve the accuracy of the methods is described. Two approaches to the search for named entities are described: using hidden Markov models and using recurrent neural networks (with LSTM and GRU cell architectures). For each method, its essence is briefly described and the results of work on test data are shown. In conclusion, the conclusion is made about the applicability of models in real work, and also an assumption is made about the promising directions of development of the described method.