Distributed network attack detection algorithm based on feature fusion

Aiming at distributed denial of service (DDoS) attacks in complex network environments, this paper proposes a DDoS attack detection model based on stack AutoEncoder, which combines convolutional neural networks (CNN) and attention mechanism. The model first inputs the input feature vectors into CNN and attention based feature extraction channels to extract local and global features, then uses an AutoEncoder for feature fusion to extract more abstract global features, and finally classifies them through a neural network. The detection accuracy on datasets CICIDS2078, CICIDS2018 and CICDDoS2019 reached 99.57%, 99.57% and 99.62%, respectively.