Sergey Volkov’s research while affiliated with Peoples' Friendship University of Russia and other places

Social media provide plenty of textual information in various languages. This information can contain or provoke hatred towards different social or religious groups. In this paper, we study methods to process short text messages in English, Hindi, and Russian and identify such intolerance with cross-lingual Transformer models. Moreover, these models can be easily adapted to analyze other languages. We fine-tuned these models with several training techniques to build accurate hate speech detectors that are robust to adversarial attacks. Additional preprocessing was carried out for all datasets to improve the quality of model training. Also, for one of the training datasets, we applied the text attack algorithm that replaces some words with synonyms. For some languages, such an attack can greatly reduce the quality of the model. Experiment results show that mixing adversarial examples to a training dataset and combining deep models to randomized ensembles allows not only to reduce test error on attacked data for languages from the dataset (Hindi, Russian) but also to achieve better accuracy in other languages.

The paper presents a text mining approach to identifying and analyzing technological trajectories. The main problem addressed is the selection of documents related to a particular technology. These documents are needed to detect a trajectory of technology. The approach includes new keyword and keyphrase detection method, word2vec embeddings-based similar document search method and fuzzy logic-based methodology for revealing technology dynamics. USPTO patent database was used for experiments. The database contains more than 4.7 million documents from 1996 to 2020. Self-driving car technology was chosen as an example. The result of the experiment shows that the developed methods are useful for effective searching and analyzing information about given technologies.

This paper considers the possibility of using machine learning methods in solving the problem of intrusion detection in software-defined networks (SDN). The work is devoted to the research and development of a network attack classifier, which is a core of the intrusion detection systems. To evaluate the methods, an existing data set was used, which includes network traffic records with a several different network attack scenarios. A comparison of machine learning methods implementing neural networks on a selected data set is presented. Based on the results, it can be concluded that the task of intrusion detection in software-defined networks can be successfully solved using deep neural networks.

SDN (Software-Defined Network) is a network in which the control plane functionality is separated from the packet forwarding layer. The paper is devoted to the study of the SDN security. A comparison of neural networks with various parameters on existing dataset is presented. CSE-CIC-IDS2018 dataset [12] provided by Canadian Institute for Cybersecurity (CIC) on AWS (Amazon Web Services) was chosen. It contains of the most relevant types of network attacks. Results show that a simple neural network, such as a multi-layer perceptron, can be used to provide basic protection against most attacks. To provide more reliable protection, complex neural networks should be used. The presented LSTM-based model showed a very good result of intrusion detection.

The paper presents a text mining approach to identifying technological trajectories. The main problem addressed is the selection of documents related to a particular technology. These documents are needed to identify a trajectory of the technology. Two different methods were compared (based on word2vec and lexical-morphological and syntactic search). The aim of developed approach is to retrieve more information about a given technology and about technologies that could affect its development. We present the results of experiments on a dataset containing over 4.4 million of documents as a part of USPTO patent database. Self-driving car technology was chosen as an example. The result of the research shows that the developed methods are useful for automated information retrieval as the first stage of the analysis and identification of technological trajectories.