Multi-Agent System for Detecting and Counteracting Attacks on the Enterprise Information System

Kostiuk, Yuliia (2025) Multi-Agent System for Detecting and Counteracting Attacks on the Enterprise Information System Колективна (три і більше авторів). Estonia, Scientific Center of Innovative Research.

Текст Y_Kostiuk_ITSC_2025.pdf Download (1MB)

Анотація

Modern enterprises face growing cyber incident frequency and increasingly diverse vectors, including AI-driven and multi-vector attacks, while cloud services, IoT, and decentralised architectures strain conventional security controls. Multi-agent attack-detection-and-prevention systems (ADPSs) are proposed as a distributed defence paradigm in which autonomous components monitor and interpret heterogeneous telemetry across network, server, and workstation layers. This study aims to design a scalable and resilient multi-agent system that detects and counteracts attacks on an enterprise information system through coordinated, context-aware decision making and continuous adaptation to evolving threats. The approach specifies an agent-based architecture and formal models for agent behaviour, cooperation, and belief updating. Threat assessment integrates neural networks with fuzzy logic and Bayesian inference, enabling dynamic updating of threat models using real-time observations and historical data. System performance is assessed through operational metrics including false positive rate, belief stability, and response effectiveness. The proposed architecture supports modular deployment of specialised agents that collect and analyse distributed security signals and coordinate responses. By combining deep learning with probabilistic modelling and adaptive learning, the system is positioned to improve detection precision and mitigate limitations of traditional ADPSs, while maintaining rapid adaptability and resilience under modern enterprise conditions. A multi-agent cyber-defence platform can strengthen enterprise security by enabling distributed monitoring, cooperative analytics, and policy-aligned response selection under uncertainty. Future work should validate the approach in real enterprise deployments, benchmark against established ADPS tools, and advance explainability, adversarial robustness, and privacy-preserving learning for sensitive logs and threat-intelligence integration.

Тип елементу :	Монографія (Колективна (три і більше авторів))
Ключові слова:	enterprise information system; multi-agent system; information security; attack detection; incident response; cyber threats; neural networks; fuzzy logic; Bayesian inference; adaptive learning; SIEM integration; IT security
Типологія:	Монографії > Видані в іноземному видавництві мовами ОЕСР/ЄС
Підрозділи:	Факультет інформаційних технологій та математики > Кафедра інформаційної та кібернетичної безпеки ім. професора Володимира Бурячка
Користувач, що депонує:	Павло Миколайович Складанний
Дата внесення:	13 Лют 2026 10:22
Останні зміни:	13 Лют 2026 10:22
URI:	https://elibrary.kubg.edu.ua/id/eprint/56409

Actions (login required)

Перегляд елементу