PREDICTION OF CYBERATTACKS USING ANOMALY DETECTION ARTIFICIAL INTELLIGENCE ALGORITHMS

Abstract

In the scientific paper by V. V. Bandura, M. V. Krykhiskyi and V. I. Chudyk titled “Prediction of Cyberattacks Using Anomaly Detection Artificial Intelligence Algorithms”, a new model for effective cyberattack prediction is examined.This model employs innovative methods for analyzing and detecting anomalies. The research focuses on developing and implementing artificial intelligence algorithms to predict cyberattacks by identifying anomalies in cyberspace. In today’s digital world, where data volume and network complexity are continuously increasing, the ability to quickly and accurately identify potential threats becomes critically important. The use of artificial intelligence in this field not only automates the process of data monitoring and analysis but also ensures higher prediction accuracy.The main focus is on combining several advanced artificial intelligence technologies, such as recurrent neural networks (RNN), attention mechanisms, convolutional neural networks (CNN), bidirectional networks (Bi-RNN), and transformers.The proposed model leverages the advantages of these technologies to create a comprehensive approach to anomaly detection in large volumes of data collected from various sources. The attention mechanism allows the model to focus on the most significant parts of the data, while convolutional neural networks enhance the processing of spatial dependencies in the data. Bidirectional networks enable the analysis of data in both directions, allowing the detection of more complex patterns and correlations, and transformers ensure efficient processing of large data sequences. As a result, the proposed model demonstrates high accuracy in predicting cyberattacks and significantly reduces the number of false positives, contributing to increased cybersecurity and protection of information systems.The integration of convolutional recurrent neural networks and transformers into a single model provides a high level of adaptability to new types of threats and anomalies that continuously emerge in cyberspace. The model shows stable performance even when analyzing large volumes of real-time data, which is critically important for modern cybersecurity systems. The high accuracy and data processing speed make this model practical and reliable for detecting cyberattacks.