Blockchain -Assisted Hybrid Deep Learning-Based Secure Mechanism for Software Defined Networks.

Software Defined Network (SDN) has emerged as an autonomous concept for dynamic management and network configuration. Blockchain technology has the potential to enhance the security of SDN in many practical scenarios against cyber attacks. However, security lapses are realized when integrating the SDN with blockchain during the updation of the malicious traffic flows. Hence, in addition to the encryption mechanisms and machine learning models, adopting the deep learning model becomes crucial to provide a secure mechanism for the SDN-blockchain environment. Designing a deep learning model as the flow analyzer in the SDN-blockchain layers is critical due to the functioning of the blockchain and its security structure. In addition, deep learning models confront the overfitting and class imbalance constraints across various cyber attacks in the SDN. Thus, this paper develops the security mechanism for the SDN by dynamically selecting the network-centric nodes and applying Hybrid Deep Learning (HDL) to learn the spatial and temporal characteristics of network traffic for flow entry verification with the help of weight dispersive regularization and adaptive ownership weight schemes. In the proposed system, the HDL integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) to perform the feature extraction with the handling of overfitting and traffic flow classification with the handling of class imbalance, respectively. The proposed HDL-SDN dynamically evaluates the traffic flows by enabling the trust measurement and resolves the gaps in the NSL-KDD dataset by enriching the NSL-KDD dataset with the features influenced by the designed blockchain-assisted SDN. Thus, the experimental results demonstrate that the HDL-SDN outperforms the existing SDN security scheme, yielding 8.55% higher accuracy while testing on the enriched NSL-KDD dataset.