Hybrid Deep Clustering-Based Algorithm for Anomaly Detection in Network Attacks

The rapid development of information technologies such as the internet and big data has greatly facilitated people's work and life, but it has also led to security issues such as data breaches, online fraud, and system hijacking. Currently, in most research on deep learning-based clustering algorithms, issues such as the feature diversity of high-dimensional data, the representativeness of clustering features, and the precision of anomaly detection still need optimization. To address these shortcomings, this paper designs a Hybrid Clustering Deep Autoencoder (HCDAE) algorithm, which uses two nested deep autoencoders to learn normal data, extract the key features that are representative of the data, and then select the optimal cluster centers in the latent representation space through an adaptive cluster center filtering mechanism. This further enhances the significance and representativeness of the latent feature representations. Finally, this paper uses benchmark datasets for comparative experiments to evaluate the performance of HCDAE. The experimental results show that HCDAE generally performs better than other models and can effectively detect network attack anomalies, providing a new research approach for the field of network attack anomaly detection.