Self-Organizing Cybersecurity Systems with Formal Verification Constraints using Multi-Agent Reinforcement Learning

Duvva, Vikas (2025) Self-Organizing Cybersecurity Systems with Formal Verification Constraints using Multi-Agent Reinforcement Learning. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (379kB) | Preview

Abstract

The introduction of highly developed cyber threats is getting increasingly challenging due to issues in network security architectures, where existing intrusion systems mainly face single entity failure, lack of adaptability, and absence of formal guarantees. This research integrates Multi-Agent Reinforcement Learning with Signal Temporal Logic specifications and Byzantine Fault Tolerance (MARL-STL-BFT) for formal verification of cybersecurity threat detection. Three independent LSTM-PPO agents undergo KL divergence-based diversity loss training and are constrained by STL specifications that encode temporal security properties such as minimum detection rates and maximum false positive thresholds. The STL monitoring framework guarantees agents maintain security posture constraints during operation, providing formal verification unavailable in purely data-driven methods. The 2-of-3 majority voting consensus mechanism establishes BFT, securing operation even when one agent is compromised. Experimental validation on DARPA's intrusion detection database proves both compliance with formal STL standards and 100% threat detection while preserving agent specialization across all four action types (Allow, Block, Quarantine, Alert) through diversity-constrained training. The AWS-deployed system achieved median latency of 570ms and 100% split decisions, demonstrating both fault tolerance and adherence to formal specifications in real-time operation.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Hamza Ibrahim, Ahmed UNSPECIFIED
Subjects:	T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	31 Aug 2026 12:43
Last Modified:	31 Aug 2026 12:43
URI:	https://norma.ncirl.ie/id/eprint/9695

Actions (login required)

View Item