Search

Search Results

• CASCADE CLASSIFICATION ALGORITHM FOR DETECTING MALICIOUS SOFTWARE BY STATIC ANALYSIS

  А.V. Kozachok , А. V. Kozachok , S.S. Matovykh

  18-35

  2025-11-10

  Abstract ▼

  A study is presented on the development and experimental validation of a two-level cascading architecture for static classification of Portable Executable (PE) format executable files. The aim of the work is to reduce computing costs without compromising the quality of malware detection. At the first level of the cascade, a decision tree model is used, trained on the ten most informative features, providing a high completeness of Recall 0.990 detection with an acceptable error of 1 kind. The second level is implemented by the random forest model on forty features and is intended for clarifying classification, reaching the metrics Precision 0.988 and Recall 0.987 with an F1 measure of 0.988. The classification threshold at the first level was established empirically, taking into account the minimization of errors of the second kind, while at the second level the optimal threshold value was determined by the Juden index, which provides a balanced ratio of sensitivity and specificity. Experiments on a representative sample have shown that with a malicious traffic fraction of < 20%, the proposed cascade reduces the average analysis time of one object by 5-12% compared to the 40-feature model while maintaining comparable classification quality.
  The time limit of the cascade,  = 20.6%, is analytically derived, confirmed by empirical data. The practical significance of the work lies in the possibility of integrating the proposed algorithm into antivirus gateways and endpoint protection tools, where fast response and high completeness of detection are required during mass scanning of mostly legitimate code.

• GROUPING PREDICTORS IN COMBINED PIECEWISE LINEAR REGRESSION

  S.I. Noskov , S.V. Belyaev

  120-127

  2025-10-01

  Abstract ▼

  The article provides a brief overview of publications on the application of combined structures containing known model forms as constituent elements in mathematical modeling of complex systems. In particular, the following are considered: an algorithm for estimating parameters for creating mathematical models of dynamic systems; structured mathematical models of an oxygen electrode and biological wastewater treatment; a combined model including ion exchange between calcium and copper; a combination of non-standard finite-difference schemes and the Richardson extrapolation method to obtain numerical solutions of two models of biological systems; a mathematical formulation of the problem and a heuristic approach to optimal planning of delivery routes in a multimodal system; a mathematical model for optimizing strategic and tactical decisions in all types of biomass-based supply chains; a method for developing models of various types for elements of chemical-engineering systems taking into account various types of available information and combining these models into a single complex. Two variants of the problem statement for calculating the estimates of the parameters of a combined piecewise linear regression are formulated: with a non-empty and empty intersection of the index sets that define the composition of the independent variables in the linear and piecewise linear components of the model. It is shown that in both cases, when the sum of absolute deviations of approximation errors is selected as the loss function, both variants are reduced to linear-Boolean programming problems. Two versions of a combined piecewise linear regression model of revenue of the mining and metallurgical company Severstal are constructed. The following production volumes are used as independent variables of the model: hot-rolled, cold-rolled and galvanized sheet, sheet with another metal coating, sheet with a polymer coating, rolled products, hardware products.

• ALGORITHMIC SUPPORT OF THE INTERFACE OF MANAGEMENT OF ROBOT-HUMAN WITH THE STEADY STATE VISUAL EVOKED POTENTIALS BASED ON THE MULTIVARIATE SYNCHRONIZATION INDEX

  Y.A. Turovsky, S.S. Kharchenko, R.V. Meshcheryakov, А.О. Iskhakova, A.Y. Iskhakov

  2020-07-10

  Abstract ▼

  The aim of the study is to build human-machine control systems. The main methods for con-structing such systems, methods for isolating evoked potentials in electroencephalograms. The article presents studies of electroencephalogram signals with steady state visual evoked potentials for different frequencies of photostimulation, based on the method of multivariate synchronization index. The influence of the length of the processed window on the accuracy of recognition of the frequency of the studied signal is considered. In the course of research, the authors verify the need for pre-processing of the original signals by means of bandpass signal filtering. In addition, the possibility of using a multi-dimensional synchronization index in multi-channel mode is being considered. The result of the authors study is recommendations on the parameters used to high-light the established visual evoked potentials in the method of multivariate synchronization index. The possibility of using algorithms based on a multivariate synchronization index in real time is shown. The results obtained are of practical importance, since they can be used to build neurocomputer interfaces based on visual evoked potentials and can be further used in the for-mation of control theory of robotic systems for various purposes and in the implementation of solutions for the organization of human-machine interaction in narrow practical problems.

• BASIC APPROACHES TO EXTRACTING TEXTUAL INFORMATION (OVERVIEW)

  V.V. Kureichik, P.S. Gerasimenko

  2024-10-08

  Abstract ▼

  This article is devoted to the review of known and modern approaches, methods and algorithms of
  full-text search. A brief history of the solution of the problem of search in unstructured text data, its development
  and relevance are described. The main task of search in text data is formulated. The definition of
  the database index is given. The target function of the search information system is defined in general
  terms and possible compromise variations of its parameters when solving various applied problems are
  described. A generalized architecture of a modern search information system is given with the division of
  the search problem into two phases: the primary extraction of relevant records and their subsequent ranking
  to form the final search results. The article provides basic descriptions of the main algorithms and
  methods of full-text search, such as: search by terms (logical search), search using trees and their varieties
  (B-trees, UB-trees, tries), search based on n-grams (including search based on frequency representation),
  use of the vector space model (VSM), search based on an inverted (reverse) index, search using the apparatus of fuzzy logic and bioinspired methods. The main advantages and disadvantages of these methods
  are given, their applicability in various conditions is described, and possible methods for optimizing
  the search for text data to improve the accuracy, speed of search and efficiency of resource use are considered.
  Possible promising directions in the field of solving the problem of primary information extraction
  are presented. Some methods for determining the similarity of text records for solving the ranking
  problem based on the apparatus of fuzzy logic are given. The article touches upon the issues of increasing
  the relevance of primary extraction using artificial intelligence methods, neural networks, fuzzy logic and
  bioinspired methods, in particular methods for expanding the search query and/or expanding the processed
  text records. The influence of the boundary conditions of the search system construction on increasing
  its efficiency is described. In conclusion, the article summarizes the review and discusses the prospects
  for further development of various full-text search methods.