Digital Twin-Enabled Lightweight Attack Detection for Software-Defined Edge Networks

With the development of software-defined edge networks, network management has become more flexible and realtime. However, this advancement has also led to critical security concerns, especially when detecting attacks efficiently in resourceconstraint environments. Existing solutions often suffer from high computational load, making them unsuitable for the fast, dynamic environments of resource-constrained edge environments. To tackle this issue, we introduce a lightweight attack detection system that combines digital twins with advanced machine learning techniques. Our approach uses a stacked sparse autoencoder (ssAE) for feature extraction and reduction and a hybrid CNNGRU model for accurate attack classification. The simulation results show that our solution significantly outperforms existing models, which are ANOVA-DNN, AE-MLP and CNN-LSTM. It achieves the highest detection accuracy at 99.72% and a suitable low time-cost at 0.215 ms, providing a good balance between accuracy and speed. Moreover, it delivers the lowest computational load compared to others, which makes it ideal for deployment in real-time resource-limited environments.