BOTNET ATTACK DETECTION AND MITIGATION IN SDN USING DEEP LEARNING TECHNIQUE

Abstract

Distributed denial of service (DDoS) attacks have been around for a while, and they're still a major
problem for network security and availability. This abstract introduces a novel hybrid paradigm for
Distributed Denial of Service (DDoS) mitigation in SDN settings. It incorporates a Semi-Deep Extreme
Learning Machine (Semi-Deep ELM) with a hybrid architecture. Implementing advanced mitigation
measures is made easier with SDN's programmability and centralised control. For better DDoS detection
accuracy, the proposed hybrid model merges the semi-deep ELM method with additional mechanisms for
increased robustness and adaptability, and with labelled and unlabelled data. The hybrid framework's
utilisation of deep learning architectures and extreme learning machines increases the model's scalability
and resilience in resisting various DDoS attacks, outperforming competing models. Model complexity,
resource allocation, and interaction with the current network architecture are some of the potential issues
and concerns that are covered. In cases when labelled data is few and real-time detection is essential, the
proposed approach utilising DP-K-means clustering offers a straightforward and effective manner of
detecting DDoS attacks. This hybrid design for DDoS mitigation in SDN makes DDoS detection easier
and faster by employing the DP-KMC technique to cluster benign traffic more closely. Faster mitigation
is possible with n, ERL-AlexNet! With the ever-evolving landscape of cyber threats, the Wu-Manber
algorithm provides a practical solution to enhance network security and resilience, ensure uninterrupted
service delivery, and minimise disruptions. It lets the system adapt its mitigation strategies to different
attack patterns and network conditions, making it more resilient against DDoS attacks..