Intelligent Traffic Analysis for DDoS Mitigation in IoT-Cloud Networks Using Clustering-Assisted Deep Learning
Abstract
The quick growth of Internet of Things (IoT) devices and cloud services infrastructures made it vulnerable to DDoS attacks that pose a threat to service availability and network reliability. In order to deal with the constraints of the traditional detection systems with massive, dynamic, and diverse traffic, the proposed study introduces a new hybrid system of intelligent DDoS detection and mitigation called Density-Aware Clustering Attention Network (DACA-Net) in IoT/Cloud systems. The first step of the framework implies the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) into clusters of traffic flows according to inherent density features and for isolating abnormal patterns without prior knowledge of the number of clusters. After that, a cluster-level behavioral representation is subjected to an Adaptive Attention Block Convolutional Network (AABCN) that uses attention-directed convolutional modules to highlight discriminant traffic feature and suppress irrelevant data. Moreover, automated hyperparameter optimization is done using Domestic Cattle Optimization Algorithm (DCOA) that improves the convergence and generalization of models. Consecutive experiments on benchmark datasets of IoT and cloud traffic reveal that DACA-Net approaches are more effective than traditional machine learning and baseline deep learning models in accuracy, precision, recall, and F1-score and incur a lower cost in computational power due to cluster-level learning.