Defending Against Multifaceted Network Attacks: A Multi-Label Meta-Learning and Lorenz Chaos MTD based Security Paradigm

Defending Against Multifaceted Network Attacks: A Multi-Label Meta-Learning and Lorenz Chaos MTD based Security Paradigm

연구 분야: Networking

학회: Journal of Network and Systems Management

Securing network infrastructure against multifaceted and sophisticated attacks is a major task in the ever-changing field of cyber security. Existing defense techniques frequently struggle to keep up with the continuously evolving multi-threat landscape, making networks open to new vulnerabilities. This paper presents an integrated security framework using SDN that employs multi-label meta-learning to detect and categorize various attack vectors while leveraging Moving Target Defense (MTD) grounded in Lorenz Chaos principles to enhance network resilience. The proposed multi-label meta-learning model intelligently detects and classifies a range of network attacks with a precision greater than 95% and a low false positive rate of 4%. Additionally, the Chaos-based MTD adds intrinsic unpredictability to the network, reducing the attack success rate from 91% (in the absence of defense) to 32%, which is a significant improvement over the 40.52% success rate observed with traditional MTD methods. A reinforcement learning (RL) agent is incorporated to automate MTD configurations by dynamically selecting optimal defense strategies based on network conditions, improving adaptability, and minimizing system latency. Performance evaluations reveal the Chaos MTD’s resilience, achieving a low 8 ms average Round Trip Time (RTT) under attack, and a packet loss rate reduction to 5%, outperforming traditional MTD approaches. This framework is evaluated with a novel blended attack scenario that demonstrates high efficiency in reducing attack success rates, maintaining low packet loss, and optimizing defense against complex and evolving threats, offering a robust, scalable solution for real-time network security.