EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

MULTI-AGENT REINFORCEMENT LEARNING FOR AUTONOMOUS AND ADAPTIVE CYBER DEFENSE SYSTEMS

Dr. T.N. Srinivas Rao, Ganji Sai Kumar, Nallam Srinivasulu, Ramala Johnson

International Journal of AI Electrical Civil and Mechanical engineering · 2025

DOI: 10.64751/067ypr17 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of AI Electrical Civil and Mechanical engineering, 2025. Verfasst von Dr. T.N. Srinivas Rao und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The increasing sophistication and frequency of cyberattacks have created significant challenges for traditional cybersecurity solutions that rely on predefined rules, static signatures, and human intervention. Modern cyber threats such as advanced persistent threats (APTs), zero-day exploits, ransomware, distributed denial-ofservice (DDoS) attacks, and insider attacks require intelligent defense mechanisms capable of adapting to rapidly evolving attack strategies. To address these challenges, this study proposes a Multi-Agent Reinforcement Learning (MARL)- Based Framework for Autonomous and Adaptive Cyber Defense Systems, which leverages collaborative artificial intelligence agents to continuously learn, detect, and respond to cyber threats in dynamic network environments. The proposed framework enables autonomous decision-making and real-time threat mitigation while minimizing dependence on manual security operations. The framework consists of multiple intelligent defense agents deployed across different layers of the network infrastructure, including endpoint devices, edge gateways, servers, and cloud environments. Each agent employs reinforcement learning techniques to observe network conditions, identify suspicious activities, and select optimal defense actions based on environmental feedback. Through cooperative multi-agent learning, the system dynamically adapts to emerging attack patterns and coordinates defense strategies such as intrusion detection, anomaly analysis, access control enforcement, traffic filtering, threat containment, and automated incident response. Deep Reinforcement Learning (DRL) models combined with policy optimization mechanisms enable agents to continuously improve their decision-making capabilities and maximize overall network security performance. Experimental evaluation was conducted using benchmark cybersecurity datasets and simulated attack scenarios involving malware infections, network intrusions, DDoS attacks, and privilege escalation attempts. The results demonstrate that the proposed MARL-based cyber defense framework achieves superior threat detection accuracy, faster response times, higher attack mitigation rates, and improved adaptability compared with conventional security systems and single-agent reinforcement learning approaches. Furthermore, the framework effectively reduces false-positive alerts and enhances resilience against previously unseen threats through continuous learning and collaborative intelligence. The findings indicate that multiagent reinforcement learning can serve as a powerful foundation for next-generation autonomous cybersecurity systems capable of providing proactive, adaptive, and scalable defense against evolving cyber threats. The proposed framework contributes toward the development of intelligent cyber defense architectures that support secure and resilient digital infrastructures.

Zitieren

APA

Rao, D. T. S., Kumar, G. S., Srinivasulu, N., & Johnson, R. (2025). MULTI-AGENT REINFORCEMENT LEARNING FOR AUTONOMOUS AND ADAPTIVE CYBER DEFENSE SYSTEMS. International Journal of AI Electrical Civil and Mechanical engineering. https://doi.org/10.64751/067ypr17

Vancouver

Rao DTS, Kumar GS, Srinivasulu N, Johnson R. MULTI-AGENT REINFORCEMENT LEARNING FOR AUTONOMOUS AND ADAPTIVE CYBER DEFENSE SYSTEMS. International Journal of AI Electrical Civil and Mechanical engineering. 2025. doi:10.64751/067ypr17

BibTeX
@article{Rao2025,
  author = {Rao, Dr. T.N. Srinivas and Kumar, Ganji Sai and Srinivasulu, Nallam and Johnson, Ramala},
  title = {MULTI-AGENT REINFORCEMENT LEARNING FOR AUTONOMOUS AND ADAPTIVE CYBER DEFENSE SYSTEMS},
  journal = {International Journal of AI Electrical Civil and Mechanical engineering},
  year = {2025},
  doi = {10.64751/067ypr17}
}
Kostenlos starten →