Adaptive Real-Time Meta-Learning Framework with Hybrid AI Classifiers for Live Detection and Mitigation of DDoS Attacks
Abstract
Distributed Denial-of-Service (DDoS) attacks remain one of the most critical and persistent threats to network security, causing severe service disruptions and economic losses by overwhelming network infrastructures with malicious traffic. Traditional detection frameworks, which rely on static or pre-trained AI classifiers, often fail to identify newly emerging or stealth attack variants due to their limited adaptability and lack of real-time learning capability. To overcome these limitations, this paper proposes an Adaptive Meta-Learning Framework integrated with Hybrid AI Classifiers for intelligent, real-time DDoS detection and mitigation. The proposed framework employs meta-learning to rapidly adapt to novel attack scenarios using minimal data, thereby enhancing model generalization and resilience to zero-day threats. A hybrid ensemble architecture combining Graph Neural Networks (GNN), Convolutional Neural Networks (CNN), and Gated Recurrent Units (GRU) is developed to capture spatial, temporal, and topological characteristics of network traffic. To strengthen robustness, an adversarial training module based on Generative Adversarial Networks (GANs) is incorporated, enabling resistance against evasion and adversarial attacks. For dynamic response, a reinforcement learning-based mitigation agent autonomously selects optimal countermeasures-such as IP blocking and rate limiting-balancing network performance and service continuity. Furthermore, an edge-cloud collaborative processing layer distributes computational tasks efficiently, ensuring low latency at the edge and scalability in the cloud. Extensive experiments on benchmark datasets (CICIDS2017, CICDDoS2019) and adversarially generated data demonstrate that the proposed framework achieves superior detection accuracy (85.7%), faster adaptation speed, and improved mitigation effectiveness compared with state-of-the-art baselines. The research establishes a comprehensive, intelligent, and scalable defense mechanism capable of maintaining network integrity in evolving threat environments. This work contributes a novel synthesis of meta-learning, hybrid ensemble intelligence, and reinforcement-based mitigation for real-time, adaptive DDoS protection.