ARTIFICIAL INTELLIGENCE SYSTEM FOR IDENTIFYING ROBOT BEHAVIOR ON A WEB RESOURCE
The architectural implementation of a machine learning system for identifying a robot on a web resource by behavioral factors is considered. The article discusses how to build software architecture for a machine learning system whose task is to determine the behavior of anonymous users. Behavioral f...
Gespeichert in:
Veröffentlicht in: | Herald of Advanced Information Technology 2019-12, Vol.2 (4), p.288-297 |
---|---|
Hauptverfasser: | , , , |
Format: | Artikel |
Sprache: | eng |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The architectural implementation of a machine learning system for identifying a robot on a web resource by behavioral factors is considered. The article discusses how to build software architecture for a machine learning system whose task is to determine the behavior of anonymous users. Behavioral factors for identification are a set of factors describing various components, each of which may be characteristic of the behavior of the robot. Weka software provides a mechanism for training on designed data models describing human and robot behavior. The learning algorithm – the “method of nearest neighbours”, provides the construction of images based on the largest number of combinations of factors that describe one of the models. Data models for training are stored in a file on the hard disk in the form of matrices of feature descriptions of each of the types of behaviors. The article discusses software and algorithmic solutions that will help solve the problems of combating fraudulent clicks, spam and distributed multi-session attacks on the server, as well as reducing the level of confidence in the website for search engines. The article discusses software and algorithmic solutions that will help solve the problems of fighting click fraud, spam and DDOS attacks, as well as reducing the level of trust of a web site for search engines. Because a large number of illiquid and malicious traffic reduces search positions and reduces the TIC (thematic citation index) and PR (page rank) of the site, which reduces the profitability of the web resource. A large number of illiquid and malicious traffic reduces search positions and reduces the thematic citation index and search ranking of site pages, which leads to a decrease in the profitability of a web resource. The results of this article are the proposed behavior analysis system, a description of the technical implementation shell and a system training model. The statistics for comparing malicious traffic after connecting the system to a web site are also given. The implementation language was selected as Java. Using this system possibly allows cross-platform integration of the system, both on Linux and Windows. Data collection from the site, to determine the role of the user, is carried out using JavaScript modules located on the web resource. All data collection algorithms and user information storage periods are implemented within the framework of the European Data Protection Regulation. The system also provide |
---|---|
ISSN: | 2663-0176 2663-7731 |
DOI: | 10.15276/hait.04.2019.5 |